Trudeau and Voting Reform in Canada (new thread)

[Warren D Smith]

I may actually be able to accomplish something good with Trudeau.
I do not want to explain the details of the somewhat lucky situation
that I'm in,
but basically I have an in with Trudeau and leading Canadian lib party members
that is 80% likely to mean they'll listen to and respect my guidance.
If I play my cards right. Which I think I know how to do. And since
I'm getting help. This will involve me co-authoring some guiding
documents for them. Has to happen ASAP.

To help: post on this thread suggestions for what you think Canada new
improved voting system ought to be. There are 4 main options in my
view:

1. Stay with single winner per district system, just choose them using
a system better than plurality, e.g. score voting.

2. Go with a PR system. Several possible flavors both (A) old (year
1900 or so) and (B) new and comparatively untested, since recently
invented, but look pretty promising
at largely avoiding many of the known weaknesses of older PR methods.

3. Go with a sort of hybrid of PR and single-winner-in-district.

I welcome+need simple+good suggestions of all 4 types 1,2A,2B,3.
A literature review/trawl of, e.g, Social Choice & Welfare, Public
Choice, J.Political Economy, last 10 years on suitable voting systems,
would also be helpful in case they've
published any new good systems we hadn't noticed.


--
Warren D. Smith
http://RangeVoting.org <-- add your endorsement (by clicking
"endorse" as 1st step)

[Jameson Quinn]

I would definitely suggest something that uses limited delegation to attain simple ballots, local representation, proportional representation, and voter accountability; a la PAL representation. I have been thinking of ways to simplify PAL representation. Here's my latest proposal:

The provinces are divided up into equal-population districts ("ridings", whatever), as today. However, as you'll see below, this is mainly a trick to keep the ballot manageably simple for voters who want that, and doesn't actually limit choice for voters who don't need simplicity.

Before the election, each candidate has the opportunity to name the set of other candidates within the same party that they consider to be part of their "faction"; as well as an ordering over the other parties. So candidate A1 could say something like "My faction is {A3, A5, A7}, and then for other parties my preferences are B, C, and D." Declared party preferences can be truncated but must be strict up to truncation. Members of a candidate's party who are not named as members of their faction are automatically inserted as preferences between the faction and all other parties.

This declaration by "faction set" is intended as a compromise between two goals: that of allowing a vote to have some content that's even more specific than the party level, and that of preventing log-rolling or entrenchment of party insiders.

Ballots are different for each district. A ballot lists the local candidates, with their parties, in a reasonably large font, on the front page. There is also a "write-in" area, consisting of a set of one bubble for each party, plus some scantron bubbles for entering a candidate code. Each candidate in the province has a different code. Codes are not unique across parties — if there is a candidate 249 in party X there will likely also be a 249 in party Y. Party/code combinations are unique across provinces — there will not be a different X249 in the next province over. Large-type booklets listing all candidate codes, indexed (and color-coded) by party, last name, first name, and any other ordering somebody might want, are available at all polling locations. If touchscreen voting is used (with a voter-verified paper ballot, naturally), then of course there is no need for codes.

If a voter votes for a nonexistent candidate code, the party they chose is still counted. Voters can also deliberately vote for party only on the "write-in" line. Unlike votes for specific candidates, party-only votes are never transferred across parties.

A vote for a candidate is considered to be delegated to that candidate's declared preference order. Candidates are elected using STV. Since tied preferences will be the norm, for the purposes of deciding which of the preferences gets the transferred votes, the ties will be broken in favor of whichever candidate has the most direct (delegated) votes. 

Partial example: say that the party A candidates get 90, 80, 70, 60, 50, and 10 votes respectively; that the Droop quota is 100 votes; and that the factions are odd versus even. In this case, the first party A candidate to be eliminated will be #6, with 10 votes. #6 declared #2 and #4 as part of their faction; of those, #2 has more votes; so the 10 votes go to #2, putting them at 90. Next to be eliminated is #5, with 50 votes. Their faction was 3 and 1, and of those two 1 has more direct votes, so 1 gets the 50. That puts 1 over the threshold, with 40 votes left over. These 40 are candidate 5's, so now they go to candidate 3, putting her also over the quota, with 10 left over. Since there are no more candidates in the faction, the 10 votes go to the remaining party member with the most direct votes (#2, who had 80 direct votes, plus the 10 indirect ones from #6). This is enough to elect #2 exactly. Next, assuming this party doesn't get any spillover votes from other parties, #4 will be eliminated, and their 60 votes will go to whichever other party #4 preferred.

One special rule: no candidate may be elected if they got less than 1/3 of a quota in direct votes. This acts like a weak "party threshold", operating similar to Germany's 5% rule. Thus, it helps prevent an excessive proliferation of geographically-diffuse parties (unless those parties can successfully organize their voters to all "write-in" the same candidate, putting that candidate about the threshold, even though their votes came from many districts). This also makes the proposal more palatable to incumbent parties, without removing its basic proportional nature, even at the sub-party "faction" level.

Once the STV election is finished and all winners are chosen, each party with any winners is required to assign territory to each of the winners so that each district is in the territory of one and only representative from each party.

As I said at the start, the advantages of this system are: simple ballots (unlike STV), local representation (unlike party lists), voter accountability (unlike closed lists), proportionality (like many others), and, among systems which have proportionality, maximal consistency with a single-member district system (so that incumbents can still run in the same district where they're well-known). While it includes several innovations, the basic underlying system is STV, which is well-understood. I'm not going to claim that there aren't some wacky strategic incentives if you dig deep enough, but really all you could accomplish with strategy is to get one more of your party's seats to go to your faction, which is pretty small potatoes as far as strategy goes.

This is not identical to PAL representation as I have described it previously, but it's pretty similar. I guess you could use the same name if you wanted to. PAL stands for Proportional, Accountable, Local.

--
You received this message because you are subscribed to the Google Groups "The Center for Election Science" group.
To unsubscribe from this group and stop receiving emails from it, send an email to electionscien...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Clay Shentrup]

Warren,

Your recent Huffington Post article shows you can be remarkably warm, engaging, and insightful when you want to be.

I would advocate:

1) Proportional Score Voting (aka RRV), which can be shown pretty simple based on the fact you can do it in a Google spreadsheet like this.

https://www.youtube.com/watch?v=kaZB84uipFk

This would include using it for executive offices too, i.e. plain Score Voting.

2) Proportional Approval Voting - even simpler

https://youtu.be/Gnsgo3z8UIg

This would include using it for executive offices too, i.e. plain Approval Voting.

3) Score Voting - single-winner

4) Approval Voting - single-winner

5) Asset Voting, possibly even for single-winner elections

I actually think it could be much better than RRV, but it's so unstudied. (Hopefully we can change that some day.)

[Clay Shentrup]

There's some speculation that Trudeau won't reform the system now that the Liberals are in power.

Perhaps it would make more sense to suggest single-winner Score or Approval Voting, since they are centrist-favoring, and the Liberals are "center-right" (whereas the NDP started as a more lefty/socialist party). Then it would have an immediate advantage for the Liberals, and would thus be more palatable. But would still be of great benefit to democracy.

In other words, if a good system happens to help those currently in power, that's the best time to get it adopted.

[Bruce Gilson]

On Thu, Oct 22, 2015 at 8:48 PM, Warren D Smith <warre...@gmail.com> wrote:

​[...]

To help: post on this thread suggestions for what you think Canada new
improved voting system ought to be.  There are 4 main options in my
view:

1. Stay with single winner per district system, just choose them using
a system better than plurality, e.g. score voting.

2. Go with a PR system.  Several possible flavors both (A) old (year
1900 or so) and (B) new and comparatively untested, since recently
invented, but look pretty promising
at largely avoiding many of the known weaknesses of older PR methods.

3. Go with a sort of hybrid of PR and single-winner-in-district.

I welcome+need simple+good suggestions of all 4 types 1,2A,2B,3.
A literature review/trawl of, e.g, Social Choice & Welfare, Public
Choice, J.Political Economy, last 10 years on suitable voting systems,
would also be helpful in case they've
published any new good systems we hadn't noticed.

​I tend to like PR systems anyway, because they allow people who are in a minority in their area to still have a representative that they can consider "theirs." But in Canada, it's probably even more important, as they already have more than 2 significant parties, so that in any constituency (Canadians use the term "riding") there are likely to be adherents of two or more local-minority parties. I would not object to any of the forms of PR that I know of, but systems like STV and RRV, where a voter has to evaluate each candidate on the ballot, require relatively small consistencies, probably

 of a size to elect

 no more than 10 members. If province-wide constituencies are desired, Ontario and Quebec would certainly be  an order of magnitude too big, if each province is to elect the same number of members it elects now, but several other provinces would be significantly larger than this limit (see http://www.ctvnews.ca/politics/election/election-2015-riding-map-of-canada). If the larger provinces are subdivided into constituencies of no more than, say, 9 members, than STV or RRV would be good. (The advantage of STV is that it's a tried method, used in elections to the Irish Dail and the Australian Senate, among others. I would love to see your RRV system tried somewhere to see how it works out in practice, and if you could convince the Canadians to try it, I'd say "great." But the fact that it is a never-before-tried system might make them reluctant to experiment with it.) ​

 

[Aaron Hamlin]

It looks like Canada's upper house runs by appointments rather than elections. That's weird. In any case:

My personal thinking is that a parliament or legislature should have diversity of opinion clustered around the average opinion. Otherwise, we may as well have a single representative using a good voting method like approval or score.

Upper house: Open Party List PR. This provides the low thresholds needed for representation of minor parties. You can do a minimum requirement like 5% to keep parties from getting too small and factional. Some countries allow parties below this vote threshold to cluster their support together to get at least one person in there.

Lower House: Proportional approval voting. Use district sizes of 5-10. This method keeps simplicity within the ballot design. Also the requirement of at least five seats avoids gerrymandering issues. It also permits a low enough threshold so that minor parties get a chance. Results are also likely less chaotic than STV.

Any single-winner seat: Approval voting. This is the most likely to get adopted because of its simplicity and similarity to current ballot design. It'll give them good and simple results.

[Rob Wilson]

This is such great news!  I was hoping someone here could influence the decision making in some way.

I personally like two round approval voting, but somehow I don't think Canadians would appreciate having two elections.  That might be better for the US or France.

If proportional representation had to be used, I'd much prefer Jameson Quinn's version of asset voting.

[Bruce Gilson]

On Fri, Oct 23, 2015 at 3:06 PM, Aaron Hamlin <aaron...@electology.org> wrote:

It looks like Canada's upper house runs by appointments rather than elections. That's weird. In any case:

My personal thinking is that a parliament or legislature should have diversity of opinion clustered around the average opinion. Otherwise, we may as well have a single representative using a good voting method like approval or score.

Upper house: Open Party List PR. This provides the low thresholds needed for representation of minor parties. You can do a minimum requirement like 5% to keep parties from getting too small and factional. Some countries allow parties below this vote threshold to cluster their support together to get at least one person in there.

​Not a bad idea, in my opinion.

​

 

Lower House:

​​

Proportional approval voting. Use district sizes of 5-10. This method keeps simplicity within the ballot design. Also the requirement of at least five seats avoids gerrymandering issues. It also permits a low enough threshold so that minor parties get a chance. Results are also likely less chaotic than STV.

​Please define

"

​p

roportional approval voting.

"

I've seen this term used in a few other posts, and I've never found it defined anywhere. Is it in fact RRV with a 2-granularity?​

 

[Rob Wilson]

If you want to make the case about how complicated IRV can be, you might want to show them some of SJVoter's videos like this one in which even the poll worker can't explain how it works. You don't want to start off with reform that nobody understands.

[Warren D Smith]

Here is one concrete tentative system suggestion for Canada designed
to be somewhat like what my liberal party contacts think they want, but better.

The goal is to elect a W-member parliament, from W-T ridings, in such
a way that
A. at least one MP is elected from each riding
(hopefully the one that riding considers the best available!),
B. plus then a further T "top up" MPs are elected countrywide,
C. in such a way as to obtain or approach "proportionality."
D. The rules of the voting system in no way even mention or require
the existence of "parties," nor assume anything about the existence
or nature of "leaders" within those parties, and thus hopefully in no
way inherently screw independent candidates or nonconformist parties.
E. System is simple.
F. The ballots are "ratings style" which allows voters to provide the
most possible information, but also (if that voter desires) less.
G. It is never suboptimal strategy, for a voter to give her honestly
favorite candidate, the max rating.
H. It is never suboptimal strategy, for a voter to give her honestly
most-detested candidate, the min rating.
I. The system is immune to "candidate cloning."
J. Voting honestly appears always to be superior voting strategy,
versus not voting,
and monotonicity holds... these for the pre-top-up stage under
assumption no voters
omit scores, and under some sort of "hope for the best candidate
behavior and play out" modeling assumption about the top-up stage
(i.e. this isn't really true for the top-up stage, but in practice
hopefully mostly true).
K. The system is "precinct countable," i.e. each precinct could publish
a compact table of "subtotals," in such a way that the whole
election process could be performed using only these precinct
subtotals tables as input.


--------

In our description various numbers (0.8, 2, 1000, 3) have been
selected by me pretty
arbitrarily and could perfectly well be replaced by other numbers. I
am regarding voters
as female and candidates as male to make wording clearer.

0. [Optional.] Each candidate in each riding receives 1000 scores of 2
before voting starts.

1. Each voter rates each candidate in her riding, on an 0-to-9 numerical scale.
(Voters also allowed to leave some candidates unrated, if that voter
wishes not to express any opinion about that candidate.)

2. In each riding, the candidate with the highest average score (among
both the scores stated as his rating by the voters in his riding, and
his initial scores from step 0 that he gets for just running) wins
the MP seat from that riding.

3. At this point we have elected W-T of the W seats in Parliament,
each via the excellent
"score voting" single-winner system. Our mission now is to fill the
remaining T seats.

4. Each candidate in each riding (regardless of whether he won or lost)
is given an amount of "juice" equal to the number of voters who rated
him max, i.e. who rated him 9 if we are using 0-to-9 scale.
(But more precisely: If some voter co-equally max-rated N candidates,
they each get 1/N amount of juice from that voter.)

5. Let J denote the total amount of juice. (Which, note, equals the
total number of voters.)

6. Remove J/W juice from each candidate who has already been elected.
(If this would make that candidate now have a negative amount of
juice, instead just give him zero.)

7. Candidates are now X'd off, one by one. Always, the candidate who currently
possesses the least juice is the one who is X'd.
Each time such a candidate is X'd, he must donate all his juice to
some not-yet-Xd candidate (or candidates, plural) of his choice
somewhere in Canada
within 3 minutes, otherwise his juice vanishes. (As this process
proceeds, internet
should broadcast the amounts of juice currently possessed by all
candidates at all times.
Any candidate worried about his internet connection could pre-provide
instructions
about how to disperse his juice for example "please give all my juice
to the first candidate
on the following list eligible to accept it.")

8. Keep doing step 7 until the number of "both not-yet-elected and
not-yet-Xd" candidates, shrinks to T.

9. Those T candidates fill the remaining T seats. The parliament has
now been fully chosen.

---end.

On 10/23/15, Rob Wilson <blahf...@gmail.com> wrote:
>
>
> If you want to make the case about how complicated IRV can be, you might
> want to show them some of SJVoter's videos like this one

> <https://www.youtube.com/watch?v=b42PLX9RTag>in which even the poll worker

> can't explain how it works. You don't want to start off with reform that
> nobody understands.
>

> --
> You received this message because you are subscribed to the Google Groups
> "The Center for Election Science" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to electionscien...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>

[Jameson Quinn]

I think that Warren's proposal has a lot of good points:

-Using an MMP-like system allows an "at least 1 candidate per riding wins" guarantee.

-Using a hybrid of score voting (for the "riding" winners) and non-divisible asset voting (for the "national" members) means you get the inherent quality and expressiveness of a score result, without its strategic incentives.

-Using a purely bottom-up system, with no top-down "vote exhaustion", removes a lot of the "avoid voting for winners" burial-type incentives that are so pervasive in PR systems.

However, I think it also has some weaknesses:

-It does not actually meet the "favorite betrayal" criterion Warren claims. Say that there are two candidates A and B I trust equally to use my delegated vote, but I expect A to win the local riding and B to get eliminated quickly. If I give my votes to A, they could very well get "stuck" there until the end of the process; if I give them to B, B will get a chance to strategize and use them to move a candidate across the line from "will probably get eliminated" to "will probably end up among the winners".

-Allowing candidates to decide freely where their "juice" (delegated votes) goes opens the door to arm-twisting and log-rolling by party committees. And the process is too opaque to an ordinary voter; it's too easy to say you're pursuing one goal but actually make a move that is tailored to some other goal, because the game theory is actually pretty complex.

-Furthermore, the fact that once you transfer juice to somebody you get no further say in where it goes next makes it too easy to do "juice-laundering" that ends up transferring juice to a corrupt insider.

-I think the "transfer juice in real time" aspect is totally unworkable. Sure, modern computers are easily up to the job; but there are so many ways that real life could get in the way.

So, here's a counter-proposal, where I try to keep what I see as the good aspects, while improving on the bad aspects (not surprisingly, using ideas I've been polishing in the context of PAL).

I keep all of Warren's desiderata, except for D and G, which I weaken.

D is his desideratum of "the system doesn't mention the word 'party'." I agree that the system should not have any explicit role for 'leaders', and I think the system should put party candidates and independent candidates on an even footing. However, I don't think keeping self-expressed party identification out of it is a good idea. I think that many (not all!) voters like parties as a cognitive shortcut for ideological evaluation. Also, parties will inevitably end up being important after the election, in choosing a prime minister; whether or not I like that idea, changing it is way beyond the scope of any proposal like this. So, as you'll see, I allow a role for party labels, though no role for party leaders.

G is the favorite betrayal criterion. I think not only that Warren has failed to meet this, but actually that it is impossible for a proportional system to meet it. It's my suspicion that one could prove a G-S-like theorem that any Droop-proportional system will leave open the possibility of strategies which risks causing your favorite to lose in the process of capturing two seats elsewhere. So while I'd definitely like a system without any obvious general strategies, I am not going to claim that mine actually meets an ironclad proportional FBC.

So, here's the system.

0. Candidates qualify to be on the ballot in some manner which is not too onerous for a serious candidate, but which keeps the number of candidates from being too high a multiple of the number of seats.

1. Each candidate freely chooses a self-identified party label. Any who choose "independent" will be considered to be alone in their own party.

2. Each candidate must make two pre-declarations, which information will be available to voters at the time of voting:

2a: Which candidates with the same party label (except for "independent") they consider to be "close allies".

2b: Which candidates with the same party label they consider to be "unacceptable" or "not really part of this party".

2c: A preference ordering over the other party labels and independent candidates, possibly truncated at the bottom.

3. [Optional.] Each candidate in each riding receives 1000 scores of 2
before voting starts.

4. Each voter rates each candidate in their riding, on an 0-to-9 numerical scale. They may leave some candidates blank. Any voter who does not explicitly rate any candidates at 0 is considered to have given a 0 to any candidates they left blank. Any voter who did rate at least 1 candidate explicitly at 0 is considered to have expressed no opinion about any candidates they left blank. (I think this rule is the safest way to interpret "intent of the voter".)

5. Voters may also explicitly designate their favorite candidate among all candidates in all ridings. This is optional. A voter who rates only one candidate at 9 is considered to have favorited that candidate. Ratings for a favorite candidate are always considered to be at least equal to the highest rating a voter gave, even if the voter explicitly rated that candidate lower. In order to do so, they will use some unambiguous coding of all qualified candidates which clearly includes party ID. If they make an invalid vote which clearly includes a party ID but no valid candidate code associated with that party, the party id will be respected.

6. In each riding, the candidate with the highest average score (among
both the scores stated as his rating by the voters in their riding, and
his initial scores from step 3 that she gets for just running, but NOT including any "favorite" scores from voters outside the riding)   wins

the MP seat from that riding.

7. At this point we have elected W-T of the W seats in Parliament,

each via the excellent "score voting" single-winner system.  Our mission now is to fill the remaining T seats.

8. Each candidate in each riding (regardless of whether they won or lost)
is considered to have a number of delegated votes equal to the number of voters who "favorited" them, plus 1/N delegated votes for each voter who co-equally rated them a 9 along with a larger set of N total candidates.

9. Let D denote the total number of delegated votes.  (Which, note, equals the

total number of voters, except those who neither chose a favorite nor rated any candidates at 9.)

10. Remove D/W delegated votes from each candidate who has already been elected.
(If this would make that candidate now have a negative number of delegated votes, instead just give them zero.)

11. Find R, the total number of delegated votes remaining after step 10, and Q=S/(T+1), the number of delegated votes sufficient to ensure election. (Recall that at this point T is the number of seats still to be filled.)

12. For any candidate who was elected in stage 6, in order from lowest to highest delegated votes, transfer their delegated votes according to the rules below.

13. For any candidate who has more than Q delegated votes, in order from lowest to highest delegated votes, transfer their excess delegated votes according to the rules below.

14. Calculate the "province-wide average score" (PWAS) for each candidate, by averaging the scores each candidate received within their riding, the "2"'s from step 3, and a "9" for each delegated vote a candidate got from outside their riding.

15. Candidates are now eliminated, one by one. If EACH of the non-eliminated members of a given party has Q delegated votes, then members of that party are protected from elimination. Other than that, the candidate with the lowest PWAS is the one who is eliminated. Each time such a candidate is eliminated, all the votes they currently have are transferred according to the rules below.

16. Transfer rules:

a) A vote is always transferred according to the candidate it was delegated to by the voter, not according to where it was last.

b) Votes are never transferred to any candidate who already has at least Q votes, nor to any candidate who has been eliminated.

c) Votes are transferred according to the predeclared preferences of the candidate to whom they were originally delegated. They go first to "faction members", then to "acceptable party members", then to members of the most-preferred party, then to the next party, and so on. As long as there is any valid recipient in a higher category, no votes are transferred to a lower category. If there is ever no valid recipient for a vote, the vote is exhausted. Votes for candidates who did not predeclare preferences, or for invalid candidates in valid parties, just go to members of that party and then are exhausted. 

d) When there are several valid options of where to transfer a vote, it is transferred to the candidate with the highest PWAS.

17. Keep doing step 7 until the number of "both not-yet-elected and
not-yet-eliminated" candidates, shrinks to T.

18. Those T candidates fill the remaining T seats.  The parliament has
now been fully chosen.

...

Key differences from Warren's proposal:

①. Predeclared transfers, with limited resolution to prevent logrolling. This accounts for most of the "extra complexity". I believe that in this regard, though, the simplicity of Warren's rules is a false economy; in practice, his delegation stage is actually highly complex from the points of view of administration and of game theory.

②. Party labels are used, though they're freely chosen by each qualified candidate.

③. "Vote for one and go home" voters are considered to have rated others at 0, rather than losing a large fraction of their voting power. However, voters still have the power to leave candidates unrated.

④. Voters are allowed to delegate to candidates from outside their riding. This gives a lot more freedom to voters who like a given party in general, but not that party's local candidate. 

⑤. Eliminations progress on the basis of scores from the voters, not on the basis of current delegated "juice" totals (except when doing so would violate party proportionality). 

The combination of ④ and ⑤ makes it MUCH MUCH easier for the voters to ensure that highly unpopular candidates are not elected through wheeling and dealing. Note that this also favors candidates who are near the ideological center of their riding.

In this system, I recommend that T be kept relatively small. The problem with making T too big is that step 10 would want to take candidates way past zero; when they got set back to 0, that would effectively end up giving an advantage to their party. This would mean that large parties and regionally-strong parties, which get lots of seats directly, would end up getting an implicit advantage. As long as T is relatively small (on the order of 20-25% of seats), I think this effect would not be a big deal.

Jameson 

[Markus Schulze]

Hallo,

in recent years, proportional representation has been
rejected in referendums in British Columbia (2005, 2009),
Ontario (2007), and Prince Edward Island (2005). Only in
the metropolitan areas, these proposals won a majority. An
important argument against proportional representation was
that the districts would become too large (in square miles).

Therefore, my recommendation would be to use ...

... proportional representation by the single transferable
vote with 5 to 10 seats per district in the metropolitan
areas (e.g. Toronto, Montreal).

... proportional representation by the single transferable
vote with 2 to 7 seats per district in the urban areas
(e.g. Calgary, Edmonton, Vancouver).

... Condorcet voting in single-winner districts in the
rural areas.

Markus Schulze

[Jameson Quinn]

Hmmm... I've just thought of a defect with both Warren's system and my modification, which is worse in mine. The problem is that proportionality is broken by the zero floor in his step 6 / my step 10. To exploit that, a voter should ensure that they are not delegating to their "riding winner". At the limit, a party whose voters all did that could get their riding seats as "freebies", not counted against the proportion they get of the extra seats. In my system, the easy way to do that would be to always delegate outside of the riding.

This is fixable in various ways. I'm not sure which is best. Here are some options:

1. If you delegate outside your riding, then any rating above 4 that you give inside the riding is counted as "no opinion" when calculating the winner of that riding. Such ratings are still used for calculating PWAS. (Pros: it's actually quite fair. Cons: it is really hard to explain why.)

2. In my step 10, you subtract D/(W+1) delegated votes total for each winner. You start out by subtracting them from the winning candidate, and when you hit 0, you continue by subtracting them proportionally across all candidates from that party. (Pros: relatively simple and easy to explain why it makes sense. Cons: could end up penalizing one party faction for the success of a totally different party faction; in the extreme, could lead to "false flag" candidates.)

3. Like 2, but don't subtract from candidates who marked the winner as "unacceptable". (Upside: prevents the "false flag" problem. Cons: beginning to get complex, and potentially exploitable. "Nope, none of us actually believe that our party leader in the safe seat is an acceptable member of our party.")

4. Count ballots that gave a high rating to their riding winner as "exhausted". (Pros: fair and hard to exploit. Cons: breaks summability, making counting and auditing much harder.)

5. Just say that if you delegate outside your riding, your vote is not counted within it. (Pros: simple. Cons: means that some voters have no say in their local representative. Also, doesn't reduce the distortions to 0 here.)

5.1 if you delegate outside your riding, your vote is counted at half weight within your riding. (Pros: Like 5, but less so. Cons: Like 5, but less so.)

6. All ratings that are not delegated are counted at half weight for finding the winner of a riding. (Pros: Like 5.1, but also helps prevent clone-type exploits within a riding. Cons: reduces the system towards approval for strategic voters and towards plurality for unstrategic voters. Easier for a nonstrategic voter to inadvertently lose voting power within their riding by rating their favorite viable candidate below top.)

7. Delegated votes outside a riding are not counted if the same party wins inside the riding (Pros: works like 4, without breaking summability. Cons: parties could get around it by having some members identify as independent.)

I can think of more ideas, but they're basically variations or combinations of the above.

Of the above, I'm leaning towards 5.1. Basically, the idea is that you shouldn't delegate outside your riding unless there are no acceptable candidates within it.

[Rob Wilson]

Personally, I think your voting system on its own is just as good or better than any other proportional system (although I'm not a big fan of PR except in municipal elections), but I don't think you'll find too many regular people who will find your method simple. 

If you have a proportional system, I'd just use asset voting  in smaller 5-ish seat districts in which STV is used after candidates get their votes.  It could be described pretty simply:

“vote for your favorite candidate.  Your candidate may donate a portion of his votes to other candidates.  Top X candidates with the most votes wins.”

Of course that leaves out the details of the vote transfer, but it drives home the point.

For a single seat, approval voting is the easiest voting system to explain.  You are trying sell reform to a people who have only used first-past-the-post and if they are anything like the Americans, I assume many of them are under the impression that there is no issue with FPTP.

We don't want to sink this because people are unfamiliar with it and are scared of even looking at it because of the math.

[Jameson Quinn]

I've been thinking further about how to fix the defect I'd identified with my proposed system. I think I've hit on something that actually makes the overall system simpler.

First off: I realized that, with my "voter intent" change, minor candidates will get plenty of lazy voters giving them default 0 ratings, so we don't need the default ratings of 2.

Second: After everyone has voted, calculate the average score for each candidate by averaging all the scores from the riding, along with one extra score of "9" for each person who delegated to that candidate in whole or in part. Then, use this average score both for finding who won each riding, and also wherever I mentioned "PWAS" in my system.

Example: say we had the following 2 districts with 2 voters each:

District 1:

voter 1: A9, B9, C0

voter 2: A0, B5, C9, delegate to D.

District 2:

Voter 3: D9, E0

Voter 4: E5, F8

Candidate averages:

A: (9+9)/2

B: (9+9+5)/3

C: (0+9)/2

D: (9+9+0+9)/4

E: (0+5)/2

F: (8)/1

With this tweak, I think this system is very good:

1. Everybody gets a "local" representative, with districts only slightly bigger than currently.

2. Anybody who wants to just "go, pick one, and go home" is getting normal voting power and is likely to be casting a vote that is close to strategically optimal.

3. No candidate is unshakably entrenched. Voters always have the power to throw somebody out if they become corrupt.

4. There are no party gatekeepers.

5. A highly-engaged voter can be sure of voting so as to give proportional voting weight to one of the viable candidates close to the desired ideological or other characteristics.

[Clay Shentrup]

On Saturday, October 24, 2015 at 7:20:46 AM UTC-7, Markus Schulze wrote:

... Condorcet voting in single-winner districts in the rural areas.

There's no reason to ever use Condorcet voting. Score and Approval Voting are radically simpler and tend to elect Condorcet winners anyway, possibly more often that ACTUAL Condorcet methods.

http://scorevoting.net/AppCW.html

[Clay Shentrup]

On Friday, October 23, 2015 at 10:50:40 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

Here is one concrete tentative system suggestion for Canada designed to be somewhat like what my liberal party contacts think they want, but better.

I'm sleep deprived right now, but I find this extremely hard to follow. I don't think it's good that a person really involved in this field has to think that hard to follow it. 

[Clay Shentrup]

On Friday, October 23, 2015 at 10:50:40 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

paraphrase: The goal is to elect a W-member parliament, from W-T ridings, with T  "top up"  MPs are elected countrywide

That would mean that the ridings all have exactly one winner. What?!

[Clay Shentrup]

On Saturday, October 24, 2015 at 10:15:45 AM UTC-7, Clay Shentrup wrote:

That would mean that the ridings all have exactly one winner. What?!

Oh, you're getting proportionality from the country-wide seats. 

[Markus Schulze]

Hallo,

> There's no reason to ever use Condorcet voting. Score
> and Approval Voting are radically simpler and tend to
> elect Condorcet winners anyway, possibly more often
> that ACTUAL Condorcet methods.

Again, as Trudeau has already promised ranked ballots
or proportional representation, there is no reason to
fight for less.

Markus Schulze

[Clay Shentrup]

On Saturday, October 24, 2015 at 5:55:05 AM UTC-7, Jameson Quinn wrote:

I don't think keeping self-expressed party identification out of it is a good idea. I think that many (not all!) voters like parties as a cognitive shortcut for ideological evaluation.

But why does the _voting system_ have to recognize that. People are free to create parties and have nominations/endorsements without any need to enshrine it in the voting algorithm.

Also, parties will inevitably end up being important after the election, in choosing a prime minister

Can you elaborate?

[Clay Shentrup]

On Saturday, October 24, 2015 at 10:20:18 AM UTC-7, Markus Schulze wrote:

Again, as Trudeau has already promised ranked ballots or proportional representation, there is no reason to fight for less.

I've heard that he *never* promised that, but merely that he would get rid of Plurality Voting. In any case, I don't know that voters care that much about such details. E.g. if you gave them a Score Voting ballot, I imagine they'd accept that, or even an Approval Voting ballot. It's like when people order "a Coke", they often just mean "some kind of brown cola soda". 

[Jameson Quinn]

2015-10-24 13:26 GMT-04:00 Clay Shentrup <cl...@electology.org>:

On Saturday, October 24, 2015 at 5:55:05 AM UTC-7, Jameson Quinn wrote:

I don't think keeping self-expressed party identification out of it is a good idea. I think that many (not all!) voters like parties as a cognitive shortcut for ideological evaluation.

But why does the _voting system_ have to recognize that. People are free to create parties and have nominations/endorsements without any need to enshrine it in the voting algorithm.

If you look at my proposal, there is absolutely no centralized power involved in parties, not even the power to keep candidates out. The only thing they do is to simplify preference orderings; instead of saying "If not elected, here's an ordered list of which of the other 100 candidates I'd give my vote to", a candidate can say "If not elected, here's the subset of my party that I like best, then here's the subset that I like OK, then here's an ordered list of the other parties." That's much, much simpler, which has several advantages. It's easier for the candidates to do; in fact, so much easier that it probably makes the difference between "impossible" and "possible". It's easier for the voters to read and appreciate. And it's harder to use to hide sneaky strategies (or, by the same token, to accidentally misuse in a way that turns out to be strategically dunderheaded). Yet it's still expressive enough to give proportionality between and even, to a large degree, within parties.

Also, parties will inevitably end up being important after the election, in choosing a prime minister

Can you elaborate?

In a parliamentary system, a prime minister and a "government" comes from a majority coalition of parties. Having too many tiny splinter parties tends to make such coalitions unstable, and can give small minorities outsized power in deciding which coalitions are successful (power which they might leverage towards their particular interests at the expense of the general interest). 

In this game of "forming a government", the parties, more than the individuals inside them, are the main unit of play. Rarely do individual representatives go against their party in this most basic of partisan coalition-building. So the voters do, and should, care about the party labels of the people they elect, above and beyond those candidates' individual ideological leanings. It is entirely legitimate to want your vote to stay within a given party until there are no acceptable candidates left in that party.

Does that make my meaning clear?

[Jameson Quinn]

Clay has said that he found Warren's proposal hard to follow. I'm sure he found my proposal even more so. We expressed them as sequences of steps, with very little explanation of why they work like that. Here's my attempt to explain them in everyday terms.

Both systems are mixed member systems. Some "local members" are elected one-per-riding; other "national members" are elected in order to balance out the proportions. The basic system for electing the local members is score voting; for the national members, it's delegated voting. 

In Warren's system, the ballot consists only of a list of local candidates. You are assumed to delegate to the candidate(s) you rate highest. The phase of assigning delegated votes is basically a free-for-all, with candidates being eliminated bottom-up at a pace of one every several minutes, and reassigning all the votes they currently hold however they see fit.

In my system, the ballot is mostly the same, but it has an additional line to allow more-engaged voters to delegate to a candidate outside the local riding. The delegation phase proceeds using the predeclared preferences of the candidates. Those preferences are stated at a level of detail that's designed to be expressive enough to respect the voters' ideological wishes, but still simple enough to be manageable and to prevent logrolling.

In Warren's system, candidates are eliminated based only on how many votes they currently hold. This can be very different from the number of votes they got from the voters initially.

In my system, candidates are eliminated starting from the ones who got the weakest scores directly from the voters. This means that there is no amount of wheeling and dealing that will get a candidate a seat if they are hated by the voters.

That's all. It's not really that complicated. Sure, stating the rules takes a lot of details, but the basic ideas are simple.

Markus Schulze has pointed out that one of the reasons previous PR proposals have been opposed in Canada was that they would have involved multi-member ridings which could get too geographically huge. In Warren's and my proposals, each riding is only a fraction bigger than in the current system — something like a quarter to a third bigger.

Jameson 

--

[Clay Shentrup]

In a parliamentary system, a prime minister and a "government" comes from a majority coalition of parties.

Why do you need a "government"? You just need "legislators". They can select their PM with Score Voting or Approval Voting. They can write and vote on legislation.

[Clay Shentrup]

I now understand all the components of Warren's system, and I think it's quite brilliant. I would just simplify it a bit, and how it's explained.

- One "local" representative is elected per riding.

- The rest of parliament is elected country-wide via proportional representation.

- Your ballot consists of a 0-5 rating for any/all candidates you wish to evaluate. You additionally mark one as your "favorite".

- The winner from each riding is the candidate with the most total points.

PHASE 1 COMPLETE

- All candidates are then awarded their "favorites"

- All unelected candidates are awarded VR/W((W-R)) extra favorites, where V=number of voters, R=number of ridings, and W=number of winners

- The candidates have an hour to redistribute their favorites.

- The W-R candidates with the most favorites after that hour is up are declared winners.

[Clay Shentrup]

On Saturday, October 24, 2015 at 11:50:27 AM UTC-7, Clay Shentrup wrote:

- All unelected candidates are awarded VR/W((W-R)) extra favorites, where V=number of voters, R=number of ridings, and W=number of winners

I meant VR/(W(W-R)).

[Clay Shentrup]

On Saturday, October 24, 2015 at 11:50:27 AM UTC-7, Clay Shentrup wrote:

The only open questions for me are:

- I award excess points to the unelected candidates, rather than subtracting from the elected candidates and risking flooring noise (e.g. raising a -3 to a 0). Does that pose more problems than it solves?

- Is it worth doing the "favorites" in order to simplify vs. calculating fractional votes. Do we really want administrators to have to say, "this ballot gives 0.3333 votes to Smith"? Is the concept of a "favorite" simpler or more intuitive? Is it worth the increased risk of ballot spoilage? I quite think so.

- Do people intuitively prefer to have candidates sequentially eliminated rather than to have the negotiation period? I prefer the negotiation period because it leaves the process more open to the "free market" of ideas. If someone just wants to give her votes to her "party" leader to redistribute on her behalf, she can. (This hopefully makes Jameson happy.) But perhaps people see that as being more corrupting, because all kinds of deals can be negotiated. Not that they can't be negotiated ahead of time with Warren's proposal, but optics are a practical issue of course.

Other than that, I don't think this system can be improved upon much. It's extremely simple and game theoretically sound.

[Jameson Quinn]

2015-10-24 15:01 GMT-04:00 Clay Shentrup <cl...@electology.org>:

On Saturday, October 24, 2015 at 11:50:27 AM UTC-7, Clay Shentrup wrote:

I now understand all the components of Warren's system, and I think it's quite brilliant. I would just simplify it a bit, and how it's explained.

- One "local" representative is elected per riding.

- The rest of parliament is elected country-wide via proportional representation.

- Your ballot consists of a 0-5 rating for any/all candidates you wish to evaluate. You additionally mark one as your "favorite".

- The winner from each riding is the candidate with the most total points.

PHASE 1 COMPLETE

- All candidates are then awarded their "favorites"

- All unelected candidates are awarded VR/W((W-R)) extra favorites, where V=number of voters, R=number of ridings, and W=number of winners

- The candidates have an hour to redistribute their favorites.

- The W-R candidates with the most favorites after that hour is up are declared winners.

The only open questions for me are:

- I award excess points to the unelected candidates, rather than subtracting from the elected candidates and risking flooring noise (e.g. raising a -3 to a 0). Does that pose more problems than it solves?

Yes. Oh God, yes.

 

- Is it worth doing the "favorites" in order to simplify vs. calculating fractional votes. Do we really want administrators to have to say, "this ballot gives 0.3333 votes to Smith"? Is the concept of a "favorite" simpler or more intuitive? Is it worth the increased risk of ballot spoilage? I quite think so.

It's not worth it if mistakes spoil the whole ballot. But in my proposal, I proposed mechanisms to degrade gracefully. If the voter tries to indicate a favorite and fails, they may still have given usable information about their favorite party. If even that is not legible/usable, then you just infer their favorite(s) from their scores.

 

- Do people intuitively prefer to have candidates sequentially eliminated rather than to have the negotiation period? I prefer the negotiation period because it leaves the process more open to the "free market" of ideas. If someone just wants to give her votes to her "party" leader to redistribute on her behalf, she can. (This hopefully makes Jameson happy.)

No, it actually doesn't. A negotiation period opens the door to candidates being elected on the strength of their capacity for "wheeling and dealing", despite being despised by the voters. Even Warren's bottom-up elimination leaves the door open for this, because he still leaves candidates choosing whom to assign their votes to on the fly. My proposal for relatively low-resolution pre-declarations, with the higher-resolution tiebreakers being provided by the will of the voters, avoids this.

 

But perhaps people see that as being more corrupting, because all kinds of deals can be negotiated. Not that they can't be negotiated ahead of time with Warren's proposal, but optics are a practical issue of course.

Other than that, I don't think this system can be improved upon much. It's extremely simple and game theoretically sound.

I agree that the basic ideas are good, but I see some real problems with your proposal, both in terms of the "extra points for unelected" idea, and in terms of the delegated free-for-all. I also think that your "just let anybody rate anyone" is a false simplicity; in terms of actual ballot layout, both Warren's and my ideas are simpler, even they require stating more rules up-front.

[Warren D Smith]

http://rangevoting.org/CanadaSA1.html

now is a web page which lays out my "tentative proposal"
in more detail and more carefully
with nice formatting.

[Clay Shentrup]

On Saturday, October 24, 2015 at 12:29:12 PM UTC-7, Jameson Quinn wrote:

- I award excess points to the unelected candidates, rather than subtracting from the elected candidates and risking flooring noise (e.g. raising a -3 to a 0). Does that pose more problems than it solves?

Yes. Oh God, yes.

Like what?

- Is it worth doing the "favorites" in order to simplify vs. calculating fractional votes. Do we really want administrators to have to say, "this ballot gives 0.3333 votes to Smith"? Is the concept of a "favorite" simpler or more intuitive? Is it worth the increased risk of ballot spoilage? I quite think so.

It's not worth it if mistakes spoil the whole ballot.

That depends on how common mistakes are. But they needn't spoil the scores.

 

If even that is not legible/usable, then you just infer their favorite(s) from their scores.

Then you add more to the algorithm.

If someone just wants to give her votes to her "party" leader to redistribute on her behalf, she can. (This hopefully makes Jameson happy.)

No, it actually doesn't.

"Doesn't" is not the opposite of "can".

A negotiation period opens the door to candidates being elected on the strength of their capacity for "wheeling and dealing"

Ability to negotiate is a strength. If someone is really despised, then she'll get fewer votes. Even the best wheeler and dealer isn't going to win if she has few votes to start with. And if she's really "despised", then of course she won't get many votes from her peers regardless of her ability to negotiate.

in terms of actual ballot layout, both Warren's and my ideas are simpler

Warren's idea is more complicated to tabulate and thus to explain to voters. Which could hurt its chances of adoption.

[Clay Shentrup]

Destroy this notion of "juice" post haste. Use something intuitive and sensible sounding like saying the candidates are awarded "votes" or "favorites".

[Bruce Gilson]

​This system has the problem that, with some districts using PR and others using a Condorcet system, the diversity of systems can be confusing. In addition, Condorcet systems suffer from the problem that if there is a cycle, the manner of settling the cycle can be difficult enough to explain that voters may never understand it.  ​

[Jameson Quinn]

2015-10-24 15:58 GMT-04:00 Clay Shentrup <cl...@electology.org>:

On Saturday, October 24, 2015 at 12:29:12 PM UTC-7, Jameson Quinn wrote:

- I award excess points to the unelected candidates, rather than subtracting from the elected candidates and risking flooring noise (e.g. raising a -3 to a 0). Does that pose more problems than it solves?

Yes. Oh God, yes.

Like what?

The "puppy-kicking" party has 20 candidates with 1 vote each (their own). They pool their "awarded" votes and elect one member. 

- Is it worth doing the "favorites" in order to simplify vs. calculating fractional votes. Do we really want administrators to have to say, "this ballot gives 0.3333 votes to Smith"? Is the concept of a "favorite" simpler or more intuitive? Is it worth the increased risk of ballot spoilage? I quite think so.

It's not worth it if mistakes spoil the whole ballot.

That depends on how common mistakes are. But they needn't spoil the scores.

 

If even that is not legible/usable, then you just infer their favorite(s) from their scores.

Then you add more to the algorithm.

Um... I think we're misunderstanding each other here. I'm arguing it's worth it to allow voters to explicitly "favorite" / "delegate", because any downsides in terms of spoiled ballots are avoidable.

If someone just wants to give her votes to her "party" leader to redistribute on her behalf, she can. (This hopefully makes Jameson happy.)

No, it actually doesn't.

"Doesn't" is not the opposite of "can".

It doesn't make me happy, is what I meant. 

A negotiation period opens the door to candidates being elected on the strength of their capacity for "wheeling and dealing"

Ability to negotiate is a strength. If someone is really despised, then she'll get fewer votes. Even the best wheeler and dealer isn't going to win if she has few votes to start with. And if she's really "despised", then of course she won't get many votes from her peers regardless of her ability to negotiate.

in terms of actual ballot layout, both Warren's and my ideas are simpler

Warren's idea is more complicated to tabulate and thus to explain to voters. Which could hurt its chances of adoption.

You propose letting voters rate any candidate anywhere in the province. Warren proposes letting them rate any candidate in the riding. I propose letting them rate any candidate in the riding, and also optionally delegate to some candidate elsewhere in the province. In terms of explanation, Warren's idea is the simplest in this regard, then yours, then mine. In terms of the task for a diligent voter, Warren's is the simplest, then mine, then yours. In terms of the task for a lazy voter, mine is the simplest, then yours and Warren's are arguably equal. 

Insofar as my proposal is more complex in some ways, I think that its advantages outweigh that.

[Clay Shentrup]

On Saturday, October 24, 2015 at 11:20:21 AM UTC-7, Jameson Quinn wrote:

In my system, the ballot is mostly the same, but it has an additional line to allow more-engaged voters to delegate to a candidate outside the local riding.

Sounds significantly more complicated.

 

The delegation phase proceeds using the predeclared preferences of the candidates.

That sounds nice in theory, but it is very challenging to create a system for clearly specifying that. Your proposal is quite complicated, and has elements which many voters would perceive as arbitrary:

2. Each candidate must make two pre-declarations, which information will be available to voters at the time of voting:

2a: Which candidates with the same party label (except for "independent") they consider to be "close allies".

2b: Which candidates with the same party label they consider to be "unacceptable" or "not really part of this party".

2c: A preference ordering over the other party labels and independent candidates, possibly truncated at the bottom.

Wow!

Warren's system allows parties to use whatever innovative system they want. Party members can use a system like yours if they want to. The government won't hold them to it, but surely the parties can do that through threat of discipline. Parties who think that's insanely over complicated can use other simpler ideas, or the candidates can just be independent and award their votes as they see fit.

In Warren's system, candidates are eliminated based only on how many votes they currently hold. This can be very different from the number of votes they got from the voters initially.

With good reason! The same reason that Plurality Voting picked a terrible outcome in Maine, for example.

The biggest problem is that your system is just incredibly complex.

E.g. "Calculate the "province-wide average score" (PWAS) for each candidate"

That will never be adopted.

[Jameson Quinn]

Have you ever dealt with pedantic teenagers? If you tell them something simple, they invent a complicated way to get around it. If you clearly foresee the various contingencies, your life and theirs ends up simpler. Same thing happens when you're programming a user interface.

My proposal on how candidates can pre-declare preferences is, as you say, a lot of words. But the resulting declarations are simple and accessible to the voters. Your proposal for a candidate free-for-all is easy to specify. But I believe that the resulting behavior and outcomes will be complex and opaque to the voters.

--

[Warren D Smith]

To Jameson Quinn:

Hi, I was trying to look at Jameson's counterproposal.

It's too complicated. Jameson, you can argue all you want about how
that wasn't "real" complexity, but MPs who have to talk to a town
meeting , write laws to be read by high school dropouts who work in
election agencies, etc, don't give a damn: it is just
too damn complex and you nitpicking about "real" versus "mere
apparent" complexity only makes that worse, not better. You've got to
simplify it, or forget about it. Even my thing may be too complex.

However, I liked some of the points you made. I even changed mine (on
the web page I made CanadaSA1.html ) to steal one of your ideas.

Indeed, I'm not actually sure I understand your proposal.., thought I
did, but then you changed it, and then I was losing track of what the
current proposal was. I tried to make a web page for your proposal
CanadaSA2.html too, until I saw you'd dropped it. If you want,
please edit that page and send me the new HTML, and I'll repost it
with your latest.

One of your ideas -- that voters in riding 57, could delegate their
"juice" votes
to a candidate from riding 103 if they wanted -- that's an idea I like
in principle.
But in practice, I hate it, because it'd be a logistical nightmare
with over 1000
candidates on ballot. Or do it via write-ins... which'd be another
total nightmare.

Your idea with party names has the potential to be helpful, but also
has the potential to be hurtful -- again with disaster ballot
containing 1000 parties. (Possible future Canada where each candidate
is a "party of one." Something of that ilk could happen.)

I don't mind if the candidates -- they are pros -- have to work hard
during election, but Joe Average voter has to have it easy and simple.

Your claims about me failing to get the properties I said I got...
well, the web page I made
on my proposal now clarifies that by stating properties & assumptions
more carefully.

You are right the "zeroing the negative juice" step is a way
proportionality can get violated (ok if no such zeroings happen) and
is a flaw I had papered over :(...
it may be it is best just to live with that flaw? I'm trying to
understand... are you claiming you have a fix for that?

I think some flaws you had in mind actually aren't because I assume it
will not happen (if you look at the two assumptions I used in my
proportionality guarantee on my web page)

The system I proposed has two simpler limit cases T=W (Pure asset
voting) and T=0
(pure score voting in ridings)...

[Clay Shentrup]

On Saturday, October 24, 2015 at 12:44:10 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

http://rangevoting.org/CanadaSA1.html

Side note: It would be wonderful if you'd use ScoreVoting.net and replace most occurrences of the terms in the site text. Regarding the substance:

> given that Canada's Liberal Party seems likely often to be victimized by IRV's center squeeze pathology, it may be against the Liberal party's own interests to use IRV (i.e. that Liberal MP would, by enacting his system, have risked hurting his own party as a nasty unintended side effect!).

I would make this your number one point, for reasons that should be obvious.

Do you really think averages are worth the complexity? Really? Simple sum seems far more resistant to attack from reform opponents.

Do you think a 0-9 scale is really worth the extra ballot space vs. a 0-5 scale?

> Remove J/(W+1) cups of juice from each candidate who has already won a seat. (Except: If this would cause that candidate now to have a negative amount of juice, instead just give him zero.)

Why not just _add_ votes to the unelected candidates to avoid the truncation error from the negative totals?

Lastly, the elimination process has this arbitrary 3-minute time limit. Ideally you eliminate as many rules (especially arbitrary ones) as possible. Is there a simpler way to do this?

Again: destroy the juice concept. It's such a remarkable turn-off that will diminish your credibility with average people.

[Clay Shentrup]

On Saturday, October 24, 2015 at 1:36:50 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

To Jameson Quinn:
Hi, I was trying to look at Jameson's counterproposal.
It's too complicated.  Jameson, you can argue all you want about how
that wasn't "real" complexity, but MPs who have to talk to a town
meeting , write laws to be read by high school dropouts who work in
election agencies, etc, don't give a damn: it is just
too damn complex and you nitpicking about "real" versus "mere
apparent" complexity only makes that worse, not better.  You've got to
simplify it, or forget about it.  Even my thing may be too complex.

Bingo.

[Clay Shentrup]

On Saturday, October 24, 2015 at 1:36:50 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

I don't mind if the candidates -- they are pros -- have to work hard during election, but Joe Average voter has to have it easy and simple.

I suppose that is a good counter to my "favorites" proposal.

[Jameson Quinn]

2015-10-24 16:36 GMT-04:00 Warren D Smith <warre...@gmail.com>:

To Jameson Quinn:

Hi, I was trying to look at Jameson's counterproposal.

It's too complicated.  Jameson, you can argue all you want about how
that wasn't "real" complexity, but MPs who have to talk to a town
meeting , write laws to be read by high school dropouts who work in
election agencies, etc, don't give a damn: it is just
too damn complex and you nitpicking about "real" versus "mere
apparent" complexity only makes that worse, not better.  You've got to
simplify it, or forget about it.  Even my thing may be too complex.

OK. I hear this criticism. 

I think that it's important to focus on the system as it would be presented to voters, and not get hung up on gotchas about the technical legal version. Heck: if you wrote out a plurality or approval out at a sufficiently fine level of detail, there would probably be passages that somebody could point to as being too complex. That would go double if you were looking at a first draft version of the rules.

But yes, I understand that even if rewritten in better language, and even if you take out the parts that voters never have to know, it may still be too complex. I'll see if I can't make it simpler.

However, I liked some of the points you made.  I even changed mine (on
the web page I made CanadaSA1.html ) to steal one of your ideas.

Indeed, I'm not actually sure I understand your proposal.., thought I
did, but then you changed it, and then I was losing track of what the
current proposal was.  I tried to make a web page for your proposal
CanadaSA2.html  too, until I saw you'd dropped it.

Huh? I didn't drop it. I still think it's a good base for a proposal, though it can probably be simplified and refined.

 

  If you want,
please edit that page and send me the new HTML, and I'll repost it
with your latest.

OK, I'll give it a try. 

One of your ideas -- that voters in riding 57, could delegate their
"juice" votes

I agree with Clay that the word "juice" is not helpful. Just say "delegated votes".

 

to a candidate from riding 103 if they wanted -- that's an idea I like
in principle.
But in practice, I hate it, because it'd be a logistical nightmare
with over 1000
candidates on ballot.  Or do it via write-ins... which'd be another
total nightmare.

My proposal was to do it via a write-in-like mechanism, using a list of candidate codes.

OK, the emails are flying thick and fast, so I'm going to send this now and continue my response in a later message.

 

[Jameson Quinn]

Revised version of the message I just sent, with additional material after the line of dashes:

--------------- new part of the email starts here ------------

Each candidate would have a human-readable short party name and a 3-digit unique code within the party. In order to make a write-in, you could look up the code. If you got the code wrong, the short party name would still be there, so your ballot would still be delegated to your favorite party.

 

Your idea with party names has the potential to be helpful, but also
has the potential to be hurtful -- again with disaster ballot
containing 1000 parties.  (Possible future Canada where each candidate
is a "party of one." Something of that ilk could happen.)

Anybody who is a "party of one" could be listed as an independent. In fact, the preference orders I've suggested would give enough flexibility for parties of up to 3 members to be listed as independents, and still have all their declared preferences count.

Or you could simply only allow declared affiliation with parties that already have some kind of legal existence. 

 

I don't mind if the candidates -- they are pros -- have to work hard
during election, but Joe Average voter has to have it easy and simple.

Your claims about me failing to get the properties I said I got...
well, the web page I made
on my proposal now clarifies that by stating properties & assumptions
more carefully.

You are right the "zeroing the negative juice" step is a way
proportionality can get violated (ok if no such zeroings happen) and
is a flaw I had papered over :(...
it may be it is best just to live with that flaw?  I'm trying to
understand... are you claiming you have a fix for that?

I believe it is not possible to fix this problem in all cases. (I think you could actually make a constructive G-S-like proof that would give a number of scenarios of which at least one must violate your stated property for any voting system).

However, I think it is possible to minimize this problem in various ways.

 

I think some flaws you had in mind actually aren't because I assume it
will not happen (if you look at the two assumptions I used in my
proportionality guarantee on my web page)

The system I proposed has two simpler limit cases T=W (Pure asset
voting) and T=0
(pure score voting in ridings)...

You say "pure asset voting", but it's actually a strange variant with bottom-up elimination, time limits on transfers, and irrevocable chained delegation. I think it's a mess.

I think you should be able to guarantee that a candidate who got top-rated by just 2 voters will not win. Your system cannot make this guarantee. My system can. In fact, it is highly unlikely that my system will ever elect a candidate with a raw score in the bottom quintile, and it could only happen if there was at least a quota-worth of voters whose ballots, in their most natural interpretation, clearly indicate that they prefer that candidate over the alternatives.

[Clay Shentrup]

Warren,

Does your system not distinguish between local and national candidates at all? Interesting.

My second attempt to simplify/clarify the system and especially the wording:

- One "local" representative is elected per riding.

- The rest of parliament is elected country-wide via proportional representation.

- Each voter rates each candidate in her riding, on an 0-to-5 numerical scale. Blank ratings are treated as zeroes.

- The winner from each riding is the candidate with the most total points.

- Each voter is viewed as having given 1 "vote" to her top-rated candidate. (If a voter rates more than one candidate co-equal top, then her vote is divided equally among them.)

- All unelected candidates are then awarded VR/(W(W-R)) extra votes, where V=number of voters, R=number of ridings, and W=number of winners

- Eliminate the unelected candidate with the fewest votes, and give her up to one minute to delegate her votes to an unelected un-eliminated candidate

- Repeat previous step till remaining unelected plus elected candidates equal desired number, and treat them all as the elected parliament

[Warren D Smith]

> You say "pure asset voting", but it's actually a strange variant with
> bottom-up elimination, time limits on transfers, and irrevocable chained
> delegation. I think it's a mess.

--well. Actually, the wording could be made simpler if I threw all that out
and just said "they redistribute their juice however they like, then
the T juiciest guys win."
But I figured that'd be a logistical nightmare with 1000 candidates
all negotiating
and redistributing? Sounded impossible in practice. So I had to constrain
the negotiations somehow to restore sanity.

[Jameson Quinn]

My proposal in a nutshell:

Candidates predeclare preferences, by dividing their own party into up to three groups (Preferred,Acceptable,Unacceptable) and giving a preferential order for any other parties they like.

Voters get a ballot with a list of local candidates and a write-in line. They may rate each of the local candidates 0-5 and optionally write-in a candidate from some other riding. There are mechanisms to make the write-in robust and easy to count, and to allow writing in a party instead of a candidate.

Blanks are counted as 0 if there is no explicit 0 on the ballot; otherwise, they're counted as blanks.

For each voter, find which candidates they count as "favorite". For ballots with a write-in, that's the write-in. For those without, that's any candidate(s) rated as 9.

Each candidate is given one additional rating of 9 for each ballot which counts them as favorite. (This makes it so that people who use the write-in aren't automatically getting an unfair advantage by having full power within the riding and also getting full power in the later phase.)

Each candidate's average is calculated. The highest average in each riding wins.

Each candidate's number of delegated votes is counted, dividing ballots with multiple favorites equally. Candidates who have already won in a riding get one quota worth of votes deducted from their total.

Candidates are eliminated in order of lowest average to highest. If a party ever has at least one quota worth of votes per non-eliminated candidate, all remaining candidates from that party are elected.

Votes are transferred according to the declared preferences of the candidate they were originally delegated to. If there are multiple candidates who are equal-top in those declared preferences, the votes go to whichever of those has the highest average score.

[Warren D Smith]

> My second attempt to simplify/clarify the system and especially the
> wording:
>
> - One "local" representative is elected per riding.
> - The rest of parliament is elected country-wide via proportional
> representation.
> - Each voter rates each candidate in her riding, on an 0-to-5 numerical
> scale. Blank ratings are treated as zeroes.
> - The winner from each riding is the candidate with the most total points.
> - Each voter is viewed as having given 1 "vote" to her top-rated candidate.
>
> (If a voter rates more than one candidate co-equal top, then her vote is
> divided equally among them.)
> - All unelected candidates are then awarded VR/(W(W-R)) extra votes, where
> V=number of voters, R=number of ridings, and W=number of winners

--I think you just blew it with this award, massively destroying
proportionality.

> - Eliminate the unelected candidate with the fewest votes, and give her up
> to one minute to delegate her votes to an unelected un-eliminated candidate
> - Repeat previous step till remaining unelected plus elected candidates
> equal desired number, and treat them all as the elected parliament

-well, I like that you are able to summarize and pretty nearly describe
a version of my system
in only a small amount of text.

But I think you are deluded about the "extra awards"
being a fix, actually I think that just kills it all.

Perhaps helpful is to point out that
the ratio 7:9 is NOT the same as the ratio (7+1):(9+1).

-----------

The zero-floor flaw re proportionality is, as Jameson implied, not so
easy to fix.
The fixes I thought of so far either do not work, or seem to be "cures
worse than the disease." I need to think about this more. The
questions are how best to fix it, and how serious a problem is it
anyhow -- maybe not so serious so don't try hard to fix it?

For example: If T is a large enough fraction of W, then the
zero-flooring will happen
very rarely, if ever. Meaning: not a serious problem. Of course, you do
not really want to have large T for other reasons, but it's a point.

[Clay Shentrup]

On Saturday, October 24, 2015 at 2:40:16 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

Perhaps helpful is to point out that the ratio 7:9 is NOT the same as the ratio (7+1):(9+1).

So if what should have been 7:9 was accidentally (7+1):9, then you'd want to multiply the 9 by 8/7 to get 8:10.29.

Riding winners should have gotten VR/W votes, but got X votes, so you need to multiply the unelected ballots by XW/VR.

Not perfect but better?

[Warren D Smith]

Jameson, I like your new "nutshell" except you may not have intended one bit:

"Each candidate is given one additional rating of 9 for each ballot
which counts them as favorite. (This makes it so that people who use
the write-in aren't automatically getting an unfair advantage by
having full power within the riding and also getting full power in the
later phase.)
Each candidate's average is calculated. The highest average in each
riding wins."

--Well then, my vote in riding #53 acts as a 9 which elects Hitler as
the MP for riding #103,
even though the voters in #103 dislike Hitler?

[Warren D Smith]

Re my asset-version top-up stage being a "mess"...

1. the worst problem with it (which also afflicts Jameson's "nutshell"
equally horribly)
is the ordering of the eliminations.
There will come a time when we ask "who do we eliminate next: Joe or Tom?"
And then it will be a near-tie. And then the entire country will
stall until every last chad is counted in a major lawsuit nightmare,
just to get past that point.

I.e, problem is, it is vulnerable to near-ties of this kind causing a
stall, and there are a huge number of opportunities for them to arise.

But I think there are good workarounds, e.g: the election authority
(EA) is allowed, whenever Joe and Tom are within 0.5% of each other
(pick a tolerance value), to just order them wrong in any way it
likes. Well, more precisely, it has to order them according to
whatever it thinks the vote counts are, but Joe is not allowed to sue
unless they were
more than 0.5% wrong. That may allow the EA to be a bit evil, but live with it.

Then the whole top-up stage will be unable to start until everybody is
nailed down
to better than 0.5% accuracy (or whatever it is) so that'll give all
the candidates plenty of time to scheme and plan their juice-transfer
strategies well ahead of time based on preliminary counts.

And then, no: I do not think this is a mess, despite Jameson's fears;
and I also think in future-Canada, internet technology will be better.
So, I think this may be a bit inelegant,
but it's no mess and seems perfectly livable.

[William Waugh]

Divide the ballot into two sections, "Riding" and "Province" (or "Nation" if the election is for the federal parliament).

In the Riding section, the voter gets to rate each candidate from the riding. Up to three write-ins should be allowed in addition to the ballot-qualified candidates. The voter could rate the write-ins as well as the ballot candidates. Writing in would be done with a code having up to five letters chosen by the candidate and a number assigned by a simple central information system that would treat all applications uniformly.

From among the candidates running to represent the riding, the candidate receiving the highest total score from the riding would represent the riding.

For the Province section of the ballot, the voter could write in the codes for up to 20 candidates and rate the candidates. The ballots would be tallied by RRV starting with a round where the riding representatives would be treated as having already won. Only the votes in the Province section would contribute to this tally. Voters who wish to can rate in the Province section the same candidates they rate in the Riding section, but this is not obligatory. There could be a check-off to make this easier to do.

For each section, an unrated or unmentioned candidate receives the rating at the bottom of the range, which should be stated as zero.

The same range should be used for both sections.

On Thursday, October 22, 2015 at 8:48:43 PM UTC-4, Warren D. Smith (CRV cofounder, http://RangeVoting.org) started the conversation at https://groups.google.com/forum/?fromgroups=#!topic/electionscience/g_JwRuUvwak

[Warren D Smith]

Discussion of the problem with "zero floor" messing up proportionality in
my (WDS's) scheme for Canada.

The problem arises whenever "milquetoast" MPs who are the favorite of
few, but the 2nd choice of many, win their riding's seat during the
score-voting-based stage.
Milquetoasts can then, when we subtract J/(W+1) juice from each local MP,
go negative. When we then alter those negative numbers to 0, that
damages proportionality.

How can this problem be fixed?

FIX-TRY #!:
Have all candidates pre-specify their juice-transfers by
pre-providing priority lists. Then, if you have negative juice, you
just transfer it all to your party colleagues and die. Proportionality
saved.

What's wrong with that fix? Mucho.
First, all that pre-specifying of orderings, seems too damn much
to ask from candidates.
Second, to really achieve PR, you need feedback,
you need to keep monitoring the current juice levels
so you can always juice-up the least-juicy among your party's candidates.
If you must plot all transfers in advance before election begins, you
cannot do that,
you'll mis-guess, and proportionality often will not be achieved.
Third, if you KNEW ahead of time you were likely to be a
milquetoast, then you'd say "transfer my (negative) juice NOT to my
party-pals, but
actually to rival parties!" You'd be able to game the system. Thus this
fix would actually makes things WORSE and the milquetoast party would
actually get
*extra* power awarded to it by the "fix."

So, fix#1 doesn't really work, it's a cure worse than the disease.

Speaking of which: how bad was the disease?
The disease is: milquetoast-rich parties effectively get more seats than they,
based on considering voter-favorites alone, proportionally "deserve."

But maybe if you are my 2nd favorite party, you actually do deserve some juice
from me. Those who win as compromises between opposed factions
maybe have value. So maybe this "flaw" was
not such a horrible thing.

Also, if T is small enough compared to W, then milquetoasts who go
negative, will be rare. If there are none, then proportionality is exact. If
there are few, it is near-exact.

FIX-TRY#2:
Suppose in a riding, the local MP is a milquetoast who goes negative.
Then, boost his juice up to zero, by removing juice from all
the other (losing) candidates in that riding and awarding it to him.
How mush to remove from each? Good question, there are various possible
formulas...

Then proceed normally.
We can argue this yields genuine PR, just with a different definition of "PR."
But arguably still a nice one.

[Clay Shentrup]

On Saturday, October 24, 2015 at 4:25:40 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

Also, if T is small enough compared to W, then milquetoasts who go negative, will be rare.

Don't you mean _big_ enough? I.e. if there are small number of ridings, then the odds than any riding-winner will go negative are low. 

[Jameson Quinn]

2015-10-24 18:54 GMT-04:00 Warren D Smith <warre...@gmail.com>:

Re my asset-version top-up stage being a "mess"...

1. the worst problem with it (which also afflicts Jameson's "nutshell"
equally horribly)
is the ordering of the eliminations.
There will come a time when we ask "who do we eliminate next: Joe or Tom?"
And then it will be a near-tie.  And then the entire country will
stall until every last chad is counted in a major lawsuit nightmare,
just to get past that point.

I.e, problem is, it is vulnerable to near-ties of this kind causing a
stall, and there are a huge number of opportunities for them to arise.

But I think there are good workarounds, e.g: the election authority
(EA) is allowed, whenever Joe and Tom are within 0.5% of each other
(pick a tolerance value), to just order them wrong in any way it
likes.  Well, more precisely, it has to order them according to
whatever it thinks the vote counts are, but Joe is not allowed to sue
unless they were
more than 0.5% wrong.  That may allow the EA to be a bit evil, but live with it.

Then the whole top-up stage will be unable to start until everybody is
nailed down
to better than 0.5% accuracy (or whatever it is) so that'll give all
the candidates plenty of time to scheme and plan their juice-transfer
strategies well ahead of time based on preliminary counts.

I think this is an acceptable solution to this issue.

 

And then, no:  I do not think this is a mess, despite Jameson's fears;
and I also think in future-Canada, internet technology will be better.
So, I think this may be a bit inelegant,
but it's no mess and seems perfectly livable.

Sure, I'm worried about "dog ate my internet" situations. But that's not my principal concern. My principal concern is that I have two desirable characteristics that your system does not meet:

1. Inasmuch as possible, the outcome should be a predictable function of the votes.

There is value here both in "predictable" and in "function of the votes". Surprises are bad; and the voters should be in charge.

2. There should be no way for a party insider to become so entrenched that voters cannot throw them out.

My system is designed to meet both of these. Yours is not.

By the way: here's my new "nutshell". Change from last version is in bold.

Candidates predeclare preferences, by dividing their own party into up to three groups (Preferred,Acceptable,Unacceptable) and giving a preferential order for any other parties they like.

Voters get a ballot with a list of local candidates and a write-in line. They may rate each of the local candidates 0-5 and optionally write-in a candidate from some other riding. There are mechanisms to make the write-in robust and easy to count, and to allow writing in a party instead of a candidate.

Blanks are counted as 0 if there is no explicit 0 on the ballot; otherwise, they're counted as blanks.

For each voter, find which candidates they count as "favorite". For ballots with a write-in, that's the write-in. For those without, that's any candidate(s) rated as 9.

Each candidate is given one additional rating of 9 for each ballot which counts them as favorite. (This makes it so that people who use the write-in aren't automatically getting an unfair advantage by having full power within the riding and also getting full power in the later phase.)

Each candidate's average is calculated twice, once using only votes from within the riding, and once including all votes. The highest within-riding average in each riding wins.

Each candidate's number of delegated votes is counted, dividing ballots with multiple favorites equally. Candidates who have already won in a riding get one quota worth of votes deducted from their total.

Candidates are eliminated in order of lowest average to highest. If a candidate ever has a full quota of votes, they are elected immediately, and all votes they have left over above one quota are transferred. The last votes they got are the first to be considered "left over".

Votes are transferred according to the declared preferences of the candidate they were originally delegated to. If there are multiple candidates who are equal-top in those declared preferences, the votes go to whichever of those has the highest global average score.

 

--
Warren D. Smith
http://RangeVoting.org  <-- add your endorsement (by clicking
"endorse" as 1st step)

[Warren D Smith]

-Right, I should have said "big enough."

[Jameson Quinn]

The "zero floor" problem relates pretty directly to "negative vote weight". 

I think the best solutions are:

1. ignore the problem.

OR

2. Boost negative candidates up to 0, by removing vote weight from other candidates within the same party, in the following order:

a) proportionally from all candidates that are mutually in the same party and faction as the negative candidate.

b) from all candidates in the same party, insofar as they got delegated weight from the riding of the negative candidate. (either inside or outside the riding)

c) proportionally from all candidates in the same party that did not unilaterally rate the negative candidate as unacceptable.

d) If there is still negative weight, remove from all candidates in other ridings who got delegated votes from the given riding.

e) If you still have a leftover negative, chances are very high somebody is getting away with something, but there's really nothing further you can do.

[Warren D Smith]

On 10/24/15, Jameson Quinn <jameso...@gmail.com> wrote:
> 2015-10-24 18:54 GMT-04:00 Warren D Smith <warre...@gmail.com>:
>
>> Re my asset-version top-up stage being a "mess"...

> JQ: I think this is an acceptable solution to this issue.

>> And then, no: I do not think this is a mess, despite Jameson's fears;
>> and I also think in future-Canada, internet technology will be better.
>> So, I think this may be a bit inelegant,
>> but it's no mess and seems perfectly livable.

> JQ: Sure, I'm worried about "dog ate my internet" situations.

--WDS: even then (e.g.cyberattack shuts down internet), broadcast via
TV would work
as emergency measure.

> But that's not my
> principal concern. My principal concern is that I have two desirable
> characteristics that your system does not meet:
>
> 1. Inasmuch as possible, the outcome should be a predictable function of
> the votes.

--the predictability in my scheme depends largely on the fact that in
real life, most candidates are party droids who behave pretty
predictably; and voters only tend to
tag Nixon as their "favorite" when they think they understand and can
predict him.
In principle the candidates might act in weird unpredictable ways, but
if so, then the whole ideal of "proportionality" kind of loses its
meaning anyhow, so who cares if it is lost?

> There is value here both in "predictable" and in "function of the votes".
> Surprises are bad; and the voters should be in charge.

--Asset voting winners are a function of BOTH the votes, and what the
candidates do.
Is that bad? I think both are good, and I think the best system
should be a function of both. It is a question of how much weight
should be given to the voters, and how much to the candidates. I
think the optimum is not 100:0. And not 0:100. It is somewhere in
between. By giving the candidates some weight you can try to get a
more harmonious parliament. There are some arguments of that ilk on
the Asset.html page. And I think
they are for real and are a genuine reason Asset is (in this respect)
superior to conventional PR systems which only input the votes.

For example, within US congress, the congressmen elect their leaders,
not the voters. Would you prefer the voters do that? Well, generally
speaking, i'd prefer the congressmen do it, but maybe not right now :)
[Actually if the US constitution had a clause saying "and when the
congress is unable to choose a speaker, the voters should do it"
which'd solve the USA's current problem. Somewhat.]
.

> 2. There should be no way for a party insider to become so entrenched that
> voters cannot throw them out.
>
> My system is designed to meet both of these. Yours is not.

--Not really. In my system, the only way a party insider is so
entrenched the voters cannot toss him, is if he has enough allies in
said party, who in turn have enough support from voters (or if that
insider himself has enough direct voter support within his riding).
OK? So you're fairly safe. If some insider "Nixon" becomes
recognized as corrupt,
he gets little direct voter support and he's likely tossed out. Or,
he might get support from others in his party, whom voters still
support, but if so, then there is a substantial chunk of the
electorate that wants Joe, Tom, and Harry, and those guys in turn want
Nixon in spite of his corruption scandal. In which case,
there is some legitimate reason Nixon should get elected MP.
He's serving a useful function, say enough views.

In contrast, with "closed party list" PR systems with the list-orders
decided in smoke-filled rooms, voters really would have extremely
little ability to remove Nixon.

Now let me make some further remarks.

My system is genuinely perfectly proportional despite the zero-floor
flaw, if all voters vote in plurality style. (Because; "milquetoast"
candidates then are impossible.)
Of course, if all voters did that, then the local MPs would
effectively be elected via plurality voting; not so good. But the
proportionality would be perfect, at least up to limits
caused possibly by T not being large enough.

And many previous proportionality theorems were proved under such
"racist voters assumptions" where you only prove you get perfect PR
under an assumption all voters sort of behave in the meanest, most
pro-their-party-to-hell-with-others, possible manner. OK? So my
system can make that same sort of claim, already, with no attempt
needed to "fix the flaw." You have to realize that "proportionality"
is kind of a crude thing
since always proved under some simplistic model like that. You never are going
to be able to avoid that kind of crudeness, no matter what you do.

More generally, my system also yields perfect PR if there are no
"negative milqetoast" MPs -- which happens in a much wider class of
situations than just "they are all pluralty style voters." So I'm not
too bothered by the whole "flaw."

Now I thought of a FIX#2 attempt to fix the flaw, and here's a
concrete version of it:
whenever a candidate would get negative juice, then award him just enough juice
to bring him up to zero, with that juice provided by removing fraction
F of the juice from every other candidate in his same riding, F chosen
just right to get him up to zero.

The idea behind this fix is: if you win your riding, you are the
"favorite" of that riding, even if not the favorite of all its voters
individually, so it is semi-legitimate to consider you the favorite of
its voters, sort of. So there is nothing so horrible about this juice
transfer between true-favorites, to this sort-of-favorite.

And then if you buy that, then proceed as usual from then on and we
then enjoy "perfect proportionality" in this altered sense. Which
seems fine.

OK, sounds great for fix #2.

But actually, I think fix #2 is not a good idea, because it steals
juice from the muilquetoast's rivals, to give it to him, and the whole
"flaw" that that milquetoast was getting too-high juice (because of
artificially raising him to 0) is just be AMPLIFIED by
this "fix." The opposite of helping.

So that suggests fix #3 is even better than fix #2: namely, fix #3 is
to DO NOTHING, just
stay with my original "zero the negatives" plan, with no
juice-stealing! That should be
less distortionary!

So, to conclude: fix #2 had a decent argument it was ok. But then we
had a decent argument doing nothing -- no fix -- was even better than
fix #2.

Which all proves, to me, pretty decently, that my original plan was
ok! It is not suffering,
and never was suffering, a severe flaw. Be happy,

OK, so I'm now fairly happy with the system I invented.
The top 2 criticisms Jameson came up with -- the "mess" and the "zero
floor flaw"
(which actually, I already knew about them both)
both seem ok to me as is, after deeper analysis.

That's not to say that Jameson's ideas are not good, and potentially
even better.
Or that some other ideas nobody has said yet, could not even better still.
But at least we've reached what plausibly looks like a good starting point.
[To JQ: Can you send me the latest nutshell, or web page, or whatever
so I can figure out what it currently is?]

Next, I want to say that I like the feature, with this kind of hybrid
system, that
you can tune it by choosing the value of T. Is single-winner or PR a
better way to run the world? Not clear. Both claim advantages. But
we can try to tune T to get the best trade-off for Canada. And that
may be good.

[Clay Shentrup]

I'm convinced any "fix" to the zero floor problem is not worth the complexity. If we sacrifice "proportionality" for utility (to the extent that Score Voting winners are highly optimal), this quite plausibly is an improvement. And even if not, presumably a small problem.

I would then back Warren's basic scheme, but with my change to make it use 0-5 total score rather than averages. I suppose 0-9 isn't much worse, but is less familiar, uses more ballot space, and may cause people to wonder, "Why not 0-10?"

[Clay Shentrup]

On Saturday, October 24, 2015 at 6:06:29 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

OK, so I'm now fairly happy with the system I invented.

Me too. Any thoughts on a marketable name? It seems that something like Rated MMP or Transferable MMP could be good. 

[Jameson Quinn]

2015-10-24 21:06 GMT-04:00 Warren D Smith <warre...@gmail.com>:

On 10/24/15, Jameson Quinn <jameso...@gmail.com> wrote:
> 2015-10-24 18:54 GMT-04:00 Warren D Smith <warre...@gmail.com>:
>
>> Re my asset-version top-up stage being a "mess"...
> JQ: I think this is an acceptable solution to this issue.
>> And then, no:  I do not think this is a mess, despite Jameson's fears;
>> and I also think in future-Canada, internet technology will be better.
>> So, I think this may be a bit inelegant,
>> but it's no mess and seems perfectly livable.
> JQ: Sure, I'm worried about "dog ate my internet" situations.

--WDS: even then (e.g.cyberattack shuts down internet), broadcast via
TV would work
as emergency measure.

"Dog ate my internet" wasn't shorthand for "internet stopped working globally", but rather for "some candidate happened to be out of touch at the key moment, and then later claims that it was because a tree branch took down their internet connection, and you're not really sure if they're telling the truth, but if they are it totally changes the winners, and so it leads to endless legal wrangling"... or things like that. 

> But that's not my
> principal concern. My principal concern is that I have two desirable
> characteristics that your system does not meet:
>
> 1. Inasmuch as possible, the outcome should be a predictable function of
> the votes.

--the predictability in my scheme depends largely on the fact that in
real life, most candidates are party droids who behave pretty
predictably; and voters only tend to
tag Nixon as their "favorite" when they think they understand and can
predict him.
In principle the candidates might act in weird unpredictable ways, but
if so, then the whole ideal of "proportionality" kind of loses its
meaning anyhow, so who cares if it is lost?

Sure, when we are trying to make proofs about proportionality, we make simplifying assumptions. But that doesn't mean that the systems we build should be as fragile as those assumptions.

This sounds like somebody defending the plurality winner because, if he wasn't pretty good, why did he get the most votes? I hate plurality in large part precisely because in practice it leads to a political monopoly of entrenched insiders. Why would we want to build another way for corrupt horse-traders to entrench themselves?

Your system could, in principle, elect a candidate who was rated above-bottom by a tiny handful of voters. Are you really going to try to tell me if you were a voter in that situation, that result wouldn't infuriate you? Or how about if your vote ended up getting transferred from A to B to G to M to Z who is directly opposed to everything you believe?

 

In contrast, with "closed party list" PR systems with the list-orders
decided in smoke-filled rooms, voters really would have extremely
little ability to remove Nixon.

Now let me make some further remarks.

My system is genuinely perfectly proportional despite the zero-floor
flaw, if all voters vote in plurality style.   (Because; "milquetoast"
candidates then are impossible.)
Of course, if all voters did that, then the local MPs would
effectively be elected via plurality voting; not so good. But the
proportionality would be perfect, at least up to limits
caused possibly by T not being large enough.

Perfect, except?

Your system can guarantee proportionality if all voters vote plurality-style and there are at least two total seats per riding. Those are two big "ifs".

And many previous proportionality theorems were proved under such
"racist voters assumptions" where you only prove you get perfect PR
under  an assumption all voters sort of behave in the meanest, most
pro-their-party-to-hell-with-others, possible manner.  OK?  So my
system can make that same sort of claim, already, with no attempt
needed to "fix the flaw."  You have to realize that "proportionality"
is kind of a crude thing
since always proved under some simplistic model like that.  You never are going
to be able to avoid that kind of crudeness, no matter what you do.

Again: you're confusing the flawed metric with the broader goal.

 

More generally, my system also yields perfect PR if there are no
"negative milqetoast" MPs -- which happens in a much wider class of
situations than just "they are all pluralty style voters."  So I'm not
too bothered by the whole "flaw."

Sounds like rationalization. 

Now I thought of a FIX#2 attempt to fix the flaw, and here's a
concrete version of it:
whenever a candidate would get negative juice, then award him just enough juice
to bring him up to zero, with that juice provided by removing fraction
F of the juice from every other candidate in his same riding, F chosen
just right to get him up to zero.

First: can you please stop saying "juice"?

Second: that's likely to lead to a candidate stealing votes from their direct ideological opponents in many cases. There's even possible strategies to make that more likely, and no reasonable way to defend against them.

 

The idea behind this fix is: if you win your riding, you are the
"favorite" of that riding, even if not the favorite of all its voters
individually, so it is semi-legitimate to consider you the favorite of
its voters, sort of.  So there is nothing so horrible about this juice
transfer between true-favorites, to this sort-of-favorite.

Wow. Just wow.

Hooray, you've talked yourself around to your original position. You're at least as smart as yourself. How productive.

Sarcasm aside: I think you can do better than this.

That's not to say that Jameson's ideas are not good, and potentially
even better.
Or that some other ideas nobody has said yet, could not even better still.
But at least we've reached what plausibly looks like a good starting point.
[To JQ: Can you send me the latest nutshell, or web page, or whatever
so I can figure out what it currently is?]

Candidates predeclare preferences, by dividing their own party into up to three groups (Preferred,Acceptable,Unacceptable) and giving a preferential order for any other parties they like.

Voters get a ballot with a list of local candidates and a write-in line. They may rate each of the local candidates 0-5 and optionally write-in a candidate from some other riding. There are mechanisms to make the write-in robust and easy to count, and to allow writing in a party instead of a candidate.

Blanks are counted as 0 if there is no explicit 0 on the ballot; otherwise, they're counted as blanks.

For each voter, find which candidates they count as "favorite". For ballots with a write-in, that's the write-in. For those without, that's any candidate(s) rated as 9.

Each candidate is given one additional rating of 9 for each ballot which counts them as favorite. (This makes it so that people who use the write-in aren't automatically getting an unfair advantage by having full power within the riding and also getting full power in the later phase.)

Each candidate's average is calculated. The highest average in each riding wins. [[Optional: calculate averages twice, once using only votes from within the riding, and once including all votes. In that case, use the highest within-riding average to decide riding winners.]]

Each candidate's number of delegated votes is counted, dividing ballots with multiple favorites equally. Candidates who have already won in a riding get one quota worth of votes deducted from their total.

Candidates are eliminated in order of lowest average to highest. If a candidate ever has a full quota of votes, they are elected immediately, and any votes they have left over above one quota are transferred (most recently-acquired first). 

Votes are transferred according to the declared preferences of the candidate they were originally delegated to. If there are multiple candidates who are equal-top in those declared preferences, the votes go to whichever of those has the highest global average score. 

Next, I want to say that I like the feature, with this kind of hybrid
system, that
you can tune it by choosing the value of T.  Is single-winner or PR a
better way to run the world?  Not clear.  Both claim advantages.  But
we can try to tune T to get the best trade-off for Canada.  And that
may be good.

[Warren D Smith]

> Your system could, in principle, elect a candidate who was rated
> above-bottom by a tiny handful of voters.

--yes.

> Are you really going to try to
> tell me if you were a voter in that situation, that result wouldn't
> infuriate you?

--be fine. See, first of all, ANYbody who is elected has only a small number of
voters, because each riding is small. (Each riding is like 1/300 of Canada.)
Second, there are "local" MPs and there are "global" or "at large" or
"nonlocal" MPs.
The local MPs all have good support within their riding. The global MPs
do not have ridings per se. They are there because the other MPs want
them to be there
because they find them helpful. Now if some global MP gets elected with
only a few votes from within his riding, but lots of other MPs want him, and he
restores proportionality, then who cares?
Sure maybe his riding is pissed since they did not like him, but the rest of
Canada is happy since he overall provides a useful service. Or
because Canada just
wanted to elect more Foobar-Party MPs, and they asked for it and
therefore got it.

See, this is a bit like party list PR, but the list orders (in this
approximate analogy) are being chosen by the candidates and MPs
themselves, as weighted by how much voters like them, via a
transparent process. They are not being chosen in some evil secret
way.

> Or how about if your vote ended up getting transferred from
> A to B to G to M to Z who is directly opposed to everything you believe?

--hey sucker, you cast your "favorite" vote to A, and so it's your own
damn fault
for misreading what A would do (or his for misreading B, but, whatever, the buck
stops at you).

>> My system is genuinely perfectly proportional despite the zero-floor
>> flaw, if all voters vote in plurality style. (Because; "milquetoast"
>> candidates then are impossible.)
>> Of course, if all voters did that, then the local MPs would
>> effectively be elected via plurality voting; not so good. But the
>> proportionality would be perfect, at least up to limits
>> caused possibly by T not being large enough.
>>
> Perfect, except?

--yes. And NO top-up system can avoid *that* (T not large enough) limitation.
Of course.

--Now here is further cute remark.
Actually, if all voters were to vote in APPROVAL style
(only 9s and 0s) then again my system would be perfectly proportional (up to
the usual unavoidable small-T limitation) always. Because, you can
easily prove
a theorem, that under that circumstance, the "negative guy moved to
zero" clause in my rules, automatically NEVER GETS USED!
So the "zero floor flaw" just never happens, with that kind of voters.

That's a nice point I just noticed, in favor of one of Steve Brams'
annoying suggestions, that my scheme should have said approval
everywhere, instead of score. (This is for the straight approval-ballot
version of Brams, not the "approval plus one favorite" version, for which this
trick fails.) However, I suspect the benefits derived from this, are not
enough to outweigh the harm to proportionality caused by demanding
straight approval ballot rather than the approval plus one favorite
modified ballot type.

> Your system can guarantee proportionality if all voters vote
> plurality-style and there are at least two total seats per riding. Those
> are two big "ifs".

--Why is the threshold value 2? If you have an answer, it would be nice
to make your reasoning explicit, just so we know where we stand. Good lemma.
If valid.

But on the other hand, as I said, it often works even without those two ifs.

>> And many previous proportionality theorems were proved under such
>> "racist voters assumptions" where you only prove you get perfect PR
>> under an assumption all voters sort of behave in the meanest, most
>> pro-their-party-to-hell-with-others, possible manner. OK? So my
>> system can make that same sort of claim, already, with no attempt
>> needed to "fix the flaw." You have to realize that "proportionality"
>> is kind of a crude thing
>> since always proved under some simplistic model like that. You never are
>> going
>> to be able to avoid that kind of crudeness, no matter what you do.
>>
> Again: you're confusing the flawed metric with the broader goal.

--yeah well, your vague "broader goal" is pretty undefined. Theorems
are well defined.
PR theorems always hold only in some cheesy "racism model" of some ilk.
All we always have is merely some hope that, if PR holds in the
super-racist circumstance, then we probably aren't doing too badly in
other circumstances where the PR might
get somewhat worse but other things will get better. That's how it is.
Maybe it is you who are confusing flawed metric with broader goal.

>> OK, so I'm now fairly happy with the system I invented.
>> The top 2 criticisms Jameson came up with -- the "mess" and the "zero
>> floor flaw"
>> (which actually, I already knew about them both)
>> both seem ok to me as is, after deeper analysis.
>>
> Hooray, you've talked yourself around to your original position. You're at
> least as smart as yourself. How productive.

--uh, yeah?

> Sarcasm aside: I think you can do better than this.

--well, I really *did* talk myself into liking it, and you have not here
talked me into disliking it. And see above for some more cute points
like the approval lemma. Which by the way seems to be compatible with
your vague desire for "broader goal," i.e, the PR theorem holds under
a wider class of goofy racism assumptions than you originally thought,
so it IS a broader less flawed metric, to paraphrase-regurgitate your
language.

Simplicity is pretty paramount for enaction I think, and even my
scheme may be too complex, in which case perhaps one must advise
Canada to abandon PR and just go
with, e.g. score voting within ridings, period. Simple and better than now.
Anyway what I intend to advise Canada to do, assuming I manage to
advise them of anything... is several good options, not just one, with
some discussion of their different virtues/flaws, and e.g. one will be
a PR scheme, one will be a non-PR scheme, etc.

[Warren D Smith]

WDS previously:

Now here is further cute remark.
Actually, if all voters were to vote in APPROVAL style
(only 9s and 0s) then again my system would be perfectly proportional (up to
the usual unavoidable small-T limitation) always. Because, you can
easily prove
a theorem, that under that circumstance, the "negative guy moved to
zero" clause in my rules, automatically NEVER GETS USED!
So the "zero floor flaw" just never happens, with that kind of voters.
That's a nice point I just noticed, in favor of one of Steve Brams'
annoying suggestions, that my scheme should have said approval
everywhere, instead of score. (This is for the straight approval-ballot
version of Brams, not the "approval plus one favorite" version, for which this
trick fails.) However, I suspect the benefits derived from this, are not
enough to outweigh the harm to proportionality caused by demanding
straight approval ballot rather than the approval plus one favorite
modified ballot type.

--WDS:
Sorry, I think what I just said, is wrong.
Maybe it is right if T is above some threshold, though.

[Clay Shentrup]

On Saturday, October 24, 2015 at 8:39:41 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

Anyway what I intend to advise Canada to do, assuming I manage to
advise them of anything... is several good options, not just one, with
some discussion of their different virtues/flaws, and e.g. one will be
a PR scheme, one will be a non-PR scheme, etc. 

Trudeau supposedly favors IRV to STV/MMP, PR in general. One would expect that, properly educated, he would favor Score or Approval Voting with single-winner districts. Not just because they are simpler and better than IRV, but because they'd also presumably be favorable to his center-ish Liberal Party.

Also, what about RRV with something like 8 winners per district, and NO nation-wide seats?

[Warren D Smith]

There are different speculations about what J.Trudeau favors, and
so on. Although JT might be able to ram thru whatever system he wants,
it is speculated that he would not do that because it'd make him look evil,
and that is too great a price for him to pay for too little benefit
(or some thinking
like that).

So it is speculated JT plans to create a study commission which will
recommend voting system ideas, then hopefully reach a consensus which
opposition parties also are fine with, then pass it as legislation.
While JT himself stays above the fray saying he's just not
manipulating anything (look Ma, no hands), etc.
Another path if they cannot reach such a consensus might be to put
several options to the Canadian public via referendum. However
it is speculated JT and FairVote Canada both have noticed that there
have so far been several voting-reform referenda all of which failed
to pass. Therefore they both prefer the legislative path. I would
speculate those failures probably have a lot to do with the fact
FairVote Canada keeps on proposing the same shitty set of ideas every
time, never realizing they are failing in the referenda every time
because their ideas are poor. No, it always is somebody else's fault,
somebody evil must have fooled all the voters, every time, over and
over, and if only those silly voters would finally see the light and
stop being fooled, it will pass... so they keep doing it again and
again.

Unfortunately the largest pressure group in Canada is "FairVote
Canada" (FVC; recommend you check their web site!). So if you believe
the speculation about JT trying to satisfy all with a consensus, then
you must further speculate JT is going to have to cater to FVC.
Which then would mean: Canada just does what FVC says to do, but rams it thru
legislatively so that the voters cannot reject it in a referendum like usual.

If you don't like that scenario, you need ideas that are simpler and
better and more appealing than what FVC had in mind. FVC really wants
some kind of PR, so it is speculated that any non-PR proposal will be
heavily fought by FVC, hence likely doomed.
On the other hand, non-PR systems can be considerably simpler than PR;
and a big problem in Canada is party bossism (they call it "heavy whipping" :)
with too many party droids too controlled by party leaders. A problem
with many kinds
of PR is that they can exacerbate that problem. Unfortunately the kinds of PR
that explicitly involve parties (e.g. "party list" systems) have
simplicity advantages.
I say "unfortunately" because those tend to be exactly the kinds of PR
that exacerbate party bossism and in particular with those they might
have to make it illegal or inherently hugely handicapped to run
independent, or make it illegal for MPs to switch parties...
yuk. (Such laws have been made, since felt necessary to prevent
"gaming the system", in at least some PR countries employing
explicitly-party-based PR systems. Such as New Zealand.)

[Jameson Quinn]

If "heavy whipping" is a concern, then it is especially important not to have a free-for-all delegation assignment phase. In practice, a free-for-all will soon reduce to "do what the party says, or else", which will be "pass your votes to this list of easily-whipped party cadres". The way to get representatives out from under the party, is to push them to go straight to the voters instead.

I'm not saying that my proposal can't be improved upon, or that it wouldn't be better if it were simpler. But the concerns I have, that motivated me to come up with it, are real. I think there's some good ideas in your proposal. But I think that, as it stands, it's exactly the kind of thing that would increase the problem with heavy whipping.

So. How can I make my proposal simpler? That's obvious: remove the "faction". As an engaged voter myself, I would see that as a loss. But I understand that I'm far from being the norm in that sense.

So. New proposal. This is basically "open party list score PR":

0. Get list of qualified candidates. This should be based on individual petitions, without an explicit partisan aspect.

1. Each candidate self-assigns a party label. Any label chosen by 3 or fewer qualified candidates is replaced by "independent".

2. Parties can reject candidates by a 2/3 majority vote. Any rejected candidates become independents.

3. Voters can rate all candidates in their riding, and optionally designate one candidate outside their riding to delegate to. If they do not so delegate, they are considered to have delegated to whichever candidate(s) they top-rated, if any. If there is more than one such candidate, the ballot is delegated in equal fractions 1/n to each.

4. Highest average score inside each riding wins.

5. Any candidates who didn't win in 4, but who have over 1 full quota of voters who top-rated or delegated to them, win. They can transfer any excess delegated votes to any candidate in their party who hasn't yet won. Any transferred votes count as top-ratings, and average scores are recalculated. (Technical note: no more than one candidate per riding should be allowed to win in this step; if there are several qualified, the winner is whichever has highest delegated votes. This limit will almost never need invoking.)

6. For each party, calculate how many quotas of delegated votes they have, and subtract one quota of votes from that number for every seat they've already won. This number is the "leftover delegated votes".

7. Eliminate any remaining candidates of the party (or independent candidate) with the fewest leftover delegated votes, allowing each eliminated candidate to transfer their delegated votes to another party (or independent candidate) if they choose. Continue to do this until no remaining parties have fewer than one quota of delegated votes.

8. Use a least-divisor method to assign remaining seats to parties, and give them to the candidates in that party with the highest averages.

[Jameson Quinn]

To make my proposal simpler, I can collapse steps 1-3. I think my proposals there are sensible ones, but they do not really relate to the proportionality of the system. So, here it is again:

Open Party List Score MMP:

Starting with a list of candidates, each of which has a riding and a party:

1. Voters can rate all candidates in their riding, and optionally designate one candidate outside their riding to delegate to. If they do not so delegate, they are considered to have delegated to whichever candidate(s) they top-rated, if any. If there is more than one such candidate, the ballot is delegated in equal fractions 1/n to each.

2. Highest average score inside each riding wins.

3. Any candidates who didn't win in 4, but who have over 1 full quota of voters who top-rated or delegated to them, win. They can transfer any excess delegated votes to any candidate in their party who hasn't yet won. Any transferred votes count as top-ratings, and average scores are recalculated. (Technical note: no more than one candidate per riding should be allowed to win in this step; if there are several qualified, the winner is whichever has highest delegated votes. This limit will almost never need invoking.)

4. For each party, calculate how many quotas of delegated votes they have, and subtract one quota of votes from that number for every seat they've already won. This number is the "leftover delegated votes".

5. Eliminate any remaining candidates of the party (or independent candidate) with the fewest leftover delegated votes, allowing each eliminated candidate to transfer their delegated votes to another party (or independent candidate) if they choose. Continue to do this until no remaining parties have fewer than one quota of delegated votes.

6. Use a least-divisor method on the leftover votes to assign remaining seats to parties, and give them to the remaining candidates in that party with the highest averages.

[Clay Shentrup]

I suggest we create a nice simple web site describing your ultimate proposal, Warren. Maybe even fundraise to create a video for it. Kind of like our fruit video that you hate, but with a very dry straightforward explanation of the system and its benefits—a script which you'd surely prefer. Marketing the idea is important. FairVote seems to be pretty mediocre at that part, but gets attention because they're virtually the only game in town. Kind of like FairVote USA had historically been.

[Warren D Smith]

On 10/25/15, Jameson Quinn <jameso...@gmail.com> wrote:
> If "heavy whipping" is a concern, then it is especially important not to
> have a free-for-all delegation assignment phase. In practice, a
> free-for-all will soon reduce to "do what the party says, or else", which
> will be "pass your votes to this list of easily-whipped party cadres".

--that is indeed a worry.
But it's hard to predict and is more complex than that.

First priority:
I do not want a situation where, if MPs are not in a big party, they're dead.
The ultimate cure for whipping is: a too-whipped MP says "i'm quitting
your party" (either moving to a new one, or becoming independent). If
the voting system makes
that severely disadvantageous, that's very bad.
And if the law actually makes it illegal to switch parties, that's even worse!
And such laws have been passed, because felt necessary to prevent "gaming"
explicitly-party-mentioning voting systems!! So I would recommend
being very cautious
about designing voting systems that explicitly involve parties and
party-IDs. It's
tempting because you can get more simplicity that way without having
to think as hard while designing -- but resist the temptation or at
least be very careful about it, because you easily can get sucked into
hellish unintended consequences.

Second.
I strongly suspect, as JQ says, that party members will be highly
encouraged to donate
juice to their colleagues, etc etc. And most of them would not even
need to be encouraged. Consider the US congress. Does it ever
happen, that, say, a republican
vote for a Democrat to be speaker? It probably has never happened. Even now
in a state of total chaos where the republicans are totally cracking
up and unable
to choose a speaker, and fully admit it -- they still are placing
party loyalty above the country and not one of them is having anything
to do with any democrat on this.
(Even though for the good of the country, they really should play nice
with the Dems and get them to help resolve the crisis by forming a
sort of "coalition government," to use
a term that never happens in US politics.)

On the other hand if some MP *does* decide to "traitorously" donate
juice to a rival party, then what? Is his party going to expel him,
or punish him? That could be risky for
that party. It kind of works both ways. The MP can use his juice as
leverage to try
to make his party do what he wants. Also, he could just switch
parties. Mind you,
MPs who switch parties willy-nilly might hurt themselves and get a bad
rep as unreliable.
More usefully and probably, MPs can allocate their juice WITHIN their
own party to
try to mold that party in the direction they want, and sort of to
choose their own party
leaders. Don't like being whipped by boss B? Send your juice to
rival nicer boss A.
Or threaten to. Or organize your pals to.

Now another thing is, right now in Canada, they usually have a
"majority government"
and the top party can then ram through whatever it wants, other
parties be damned.
It is rarer that Canada has a "coalition government" where parties
have to play nicer with each other. But probably in any
future-PR-Canada, coalition governments will become much more common
and parties will proliferate. In that case, the nature of whippage
will need to change, and interparty alliances will need to be fostered.

So it's pretty unclear how it'll all play out. I am optimistic about
asset voting in this respect. I think asset voting will tend to yield
more harmonious and less-gridlocked parliaments. You'll have to play
nice with the other MPs so they give you juice
when you need it. You'll have to be a proven nice and good negotiator
in the view of
your colleagues to get into parliament (or anyhow it'll help). I have
no proof of this
whole theory, but it was one of my motivations back when I invented
asset voting.

Some PR countries have got into severe trouble basically because they
gridlocked, parties would not cooperate, and they had to have a
coalition but largely couldn't --
it was like perpetual state of blackmail threats all the time by all
little factions to crash the fragile system in parliament. So
gridlock is a big risk with PR. What kinds of PR
encourage/discourage it? Hard to say. I think Italy was renowned for
having that kind of dysfunction often.

In contrast, right now in USA, with voters alone controlling it,
they'll elect some asshole like Ted Cruz, whom as far as I can tell
practically everybody in the US Senate hates since he likes doing
stuff like shutting down the whole US govt for basically no reason
other than to grandstand, and seems unwilling to compromise on
anything. He was quoting "green eggs and ham" on the senate floor.
Ted Cruz and clones I suspect would have a lot more trouble getting
elected under asset voting based PR as opposed to more conventional PR
systems, because his colleagues hate him.

The MPs actually know who is good and a useful colleague. Voters do
not know that,
Voters know what that MP tells them, and how he votes, but they do not
know what it
is like to work with him on a daily basis.

Now mainly the present concern with whippage in Canada (I'm speaking
of the parliament, not the porn industry), is just about votes on Acts
of Parliament and
Motions therein. MPs are whipped to vote the way their leaders want. Otherwise
their party punishes them and the rules of parliamentary procedure
give the whippers a lot
of power to do that. Those rules need reform but that is not directly
related to what
we are speaking of here and I can't hope to affect that and know
little about it.

> The
> way to get representatives out from under the party, is to push them to go
> straight to the voters instead.

--yes and no. It's complicated. And hard to understand. But yes,
you are thinking
about the right questions. The answers may be very difficult to know...

> I'm not saying that my proposal can't be improved upon, or that it wouldn't
> be better if it were simpler. But the concerns I have, that motivated me to
> come up with it, are real. I think there's some good ideas in your
> proposal. But I think that, as it stands, it's exactly the kind of thing
> that would increase the problem with heavy whipping.
>
> So. How can I make my proposal simpler? That's obvious: remove the
> "faction". As an engaged voter myself, I would see that as a loss. But I
> understand that I'm far from being the norm in that sense.
>
> So. New proposal. This is basically "open party list score PR":
>
> 0. Get list of qualified candidates. This should be based on individual
> petitions, without an explicit partisan aspect.
>
> 1. Each candidate self-assigns a party label. Any label chosen by 3 or
> fewer qualified candidates is replaced by "independent".
>
> 2. Parties can reject candidates by a 2/3 majority vote. Any rejected
> candidates become independents.
>
> 3. Voters can rate all candidates in their riding, and optionally designate
> one candidate outside their riding to delegate to. If they do not so
> delegate, they are considered to have delegated to whichever candidate(s)
> they top-rated, if any. If there is more than one such candidate, the
> ballot is delegated in equal fractions 1/n to each.
>
> 4. Highest average score inside each riding wins.

--and once you win, you cannot be kicked out by the rest of your process, right?

> 5. Any candidates who didn't win in 4, but who have over 1 full quota of
> voters who top-rated or delegated to them, win. They can transfer any
> excess delegated votes to any candidate in their party who hasn't yet won.

--what about independents? Some independent with a lot of excess juice should
be able to use it to win friends and influence people? No? And what is your
definition of a "quota"? And why the "who hasn't yet won"?

> Any transferred votes count as top-ratings, and average scores are

> recalculated. *(Technical note: no more than one candidate per riding

> should be allowed to win in this step; if there are several qualified, the
> winner is whichever has highest delegated votes. This limit will almost

> never need invoking.)*

--By "transferred" do you mean "transferred or delegated?"
--Highest delegated votes, or highest recalculated average?
(By the way, your term "delegated votes" is making things more not
less confusing,
and then the further adjectives "excess" and "leftover" are making it
still harder. Use an entirely unquestionably distinct word like
"juice" and that entirely unnecessary wholy avoidable completely
self-inflicted confusion is never even approached, plus have fewer
words.)

> 6. For each party, calculate how many quotas of delegated votes they have,
> and subtract one quota of votes from that number for every seat they've
> already won. This number is the "leftover delegated votes".

--which could be negative?

> 7. Eliminate any remaining candidates of the party (or independent
> candidate) with the fewest leftover delegated votes,

--you'd just said "parties" own leftover delegated votes, but now you
are speaking of
"candidates" owning them. Huh? And what does "remaining" mean?
Does it mean "unelected"? By "any" did you mean "all"?

> allowing each
> eliminated candidate to transfer their delegated votes to another party (or
> independent candidate) if they choose. Continue to do this until no
> remaining parties have fewer than one quota of delegated votes.

--"until all remaining parties have at least one quota of delegated votes"
for those who do not love double negatives.
So you are now eliminating "parties," not "candidates" like you'd just
said 1 sentence ago?
And are we speaking of leftover or regular delegated votes?

> 8. Use a least-divisor method to assign remaining seats to parties, and
> give them to the candidates in that party with the highest averages.

--that all was absurdly confusing.

However, I think I get your gist+goal: to design a
top up system based on an "open party list" system with:
A. the list orderings within each party being determined "openly"
by the voters via score voting, but in combination with first
vote-delegation, and then second with asset-like vote-transferring via
your "recalculated averages" (or perhaps some other numerical
combining formula could be devised, I don't think you have your
particular one set in stone?)
B. The number of seats each party deserves, is
determined using an asset-like juice-transfer system,
except you force all juice-transfers to be within-party, so actually
the transfers do
do not matter at all for the question of how much total seats that
party deserves...
until a later stage when you allow cross-party juice transfers?
C. Finally we use a divisor system to divvy out the seats from the
lists and the party shares.

--By the way, what the hell are you complaining about re "oh my god,
the whips will force MPs to donate only within party" and then you
actually DEMAND that? Huh?

--So what do I think of this overall plan?
Well, it is more complicated with more moving parts than my plan which
already seemed
likely too complex. Part C alone already is quite complex (even though
nice and well known to those who know it and you just toss it off in 1
sentence) and already by itself approaching (what many think are) the
limits of enactible complexity.

If your goal is REALLY to give the voters maximum control since afraid
otherwise evil whips will mess up the world, and make everything
completely a function of the votes, then the simplest thing you could
do is
simply order the list within each party by order of average scores
from the voters.
Period, full stop. ... And... see my next post (to follow)

[Warren D Smith]

A simple-to-describe PR system in which everything is a function of
the votes only... makes explicit use of parties (unfortunately)...
score-style ratings ballots
plus "name favorite party" plurality-style ballot.

A POSSIBLE SYSTEM filling W seats from R ridings:
1. Voters use score voting within ridings (i.e. each voter scores each
candidate on 0-9 scale), plus also each voter names her favorite party
nationwide.
2. Each time a voter names a party, that party gets 1 cup juice.
3. Each candidate has a score (his average score from step 1).
4. Each party deserves a number of seats proportional to the total
amount juice they own.
[Details about roundoff to integers should be supplied here.]
5. Each riding deserves either floor(W/R) or ceiling(W/R) seats.
The simplest case is W=R then all ridings get exactly 1 seat.
6. Elect the W-candidate set with highest sum of scores, subject to
constraints 4 & 5.
[This is a polytime solvable graph problem. Specifically, there is a
bipartite graph
(or multigraph). Red vertices are ridings. Blue vertices are parties.
Edges are candidates
from that party within that riding. Edges have numerical scores. The
problem here is to select the subgraph with highest sum of
edge-scores, satisfying the constraints
that every red vertex have floor(W/R)<=valency<=ceiling(W/R), and
every blue vertex has valency equalling that party's deserved
seat-count. It is known how to solve such "degree constrained
subgraph" problems optimally by polytime algorithms. In the simplest
case W=R this is actually a "max weight bipartite matching" problem.]


This system could be made a bit more friendly to independents and small parties
as follows. Regard each independent as a "party of one." In step 1,
voters instead of naming a party, name some person within that party
(he gets their cup of juice). So far, these changes have been merely semantic.
But now add a new clause just before step 4: all juice-owners are
allowed to donate
some or all of their juice to whichever other candidates they please,
nationwide,
in such a way that the total juice-shares for each party then
correspond exactly to
*integer* seat counts.

[Jameson Quinn]

The thing I am afraid of is not "candidates only transfer within parties", but rather "parties, not candidates or voters, decide list order; so candidates know they have to do exactly what the party says".

Both Clay and I have said we hate the word "juice". You say you hate the words "delegated votes". I don't understand your reasoning, but I think it has to do with the fact that my explanation was unclear in ways other than my choice of words for that concept.

You just sent another mail so I'll respond to that.

[Jameson Quinn]

2015-10-25 16:02 GMT-04:00 Warren D Smith <warre...@gmail.com>:

A simple-to-describe PR system in which everything is a function of
the votes only... makes explicit use of parties (unfortunately)...
score-style ratings ballots
plus "name favorite party" plurality-style ballot.

A POSSIBLE SYSTEM filling W seats from R ridings:
1. Voters use score voting within ridings (i.e. each voter scores each
candidate on 0-9 scale), plus also each voter names her favorite party
nationwide.
2. Each time a voter names a party, that party gets 1 cup juice.
3. Each candidate has a score (his average score from step 1).
4. Each party deserves a number of seats proportional to the total
amount juice they own.
[Details about roundoff to integers should be supplied here.]
5. Each riding deserves either floor(W/R) or ceiling(W/R) seats.
The simplest case is W=R then all ridings get exactly 1 seat.

W=R is just Balinski's "fair majority voting" with score ballots. In my opinion, this is a pretty good system overall. But I think that people would feel it was unfair. It would be very hard to explain to somebody who voted for candidate A in a given riding why candidate B who got a lower score average (or total, or whatever you're using) ended up winning. I think that's a deal-breaker for that system.

OK, again, recall that both Clay and I have said that we very strongly think using the word "juice" makes us look non-serious. So I'll use "delegated votes". It means exactly the same thing. Do you have a compromise proposal?

Your system, with clause 3.5, is very similar to my latest proposal. The two differences are:

1. In my system, the only delegated votes that can be transferred across parties are the non-integer remainders, in order from lowest remainder to highest remainder. For independents, that's all there ever is, so that doesn't change anything for them versus what you have. When it is a party's turn to transfer their remainder, the non-winning members of that party decide which party the remainder goes to, in proportion to the number of delegated votes they hold.

2. In my system, if a single candidate gets over 1 quota of delegated votes, they may use the excess to help move other candidates in their party (or, if they are an independent, in other parties) up in the "list order". This works by adding more top-ratings into the average for the recipient; so this power is probably not strong enough to save a candidate that is truly disliked by the voters.

Example:

There are 20 seats and 15 ridings.

Party A gets a total of 8.4 Droop quotas of delegated votes, and wins in 6 ridings. None of their candidates gets more than a quota individually. The average scores of their non-winning candidates A1-A6 are 6, 5, 4, 3, 3, 3.

Party B gets a total of 4.6 Droop quotas of delegated votes, and wins in 3 ridings. One of those district winners, B1 had 1.2 quotas individually. The average scores of their non-winning candidates B2-B7 are also 6, 5, 4, 3, 3, 3.

Party C gets a total of 4.5 Droop quotas of delegated votes, and wins in 5 ridings. None of their candidates gets more than a quota individually. The average scores of their non-winning candidates C1-C6 are 6, 5, 4, 3, 3, 3.

Party D gets a total of 1.5 Droop quotas of delegated votes, and wins in 0 ridings. The average scores of their candidates D1-D3 are 5, 4, 3.

Independent candidate E1 gets a total of 1.3 Droop quotas and wins their riding.

Party F gets a total of 0.7 Droop quotas. Their candidates F1-F3 average 7, 4, 3.

So first, the individuals with excess delegated votes get to transfer them. B1 must give her 0.2 quotas to somebody in their party; they choose B3, bringing B3's average score from 5 up to 6.5. E1 can give his 0.3 quotas to anybody in any party. He gives to D2, bringing party D up to 1.8 quotas, and D2's average up from 4 to 6.

Next, party seats are distributed. A has 2 extra quotas, so A1 and A2 win. B has 1 extra quota, so B3 wins (average 6.5). C has negative 0.5 extra quotas, there's nothing to be done about that. D has 1 extra quota, so D2 wins (thanks to E's help).

There are 2 seats left. The remainders are now: A:0.4, B:0.6, D:0.8, F:0.7. Party A has the least, so transfers first. A3 controls half of those (0.2), and gives them to party B. A4-A6 control the other 0.2, and all give them to party D, bringing them up to 1 quota.

Now B gets to transfer. 0.6 votes. Party B are hard-liners and all decide to let their votes die rather than give them to another party.

Now there are 2 seats left, and D and F have 1 and 0.7 quotas respectively. A divisor system is used to see the obvious, that each of them gets 1 of the seats; D1 and F1 are elected.

Jameson

[Warren D Smith]

> OK, again, recall that both Clay and I have said that we very strongly
> think using the word "juice" makes us look non-serious.

--ok, again, I very strongly feel that fools who turn away from
simple, easily comprehensible natural words in favor of confusing
multiword phrases that are wholy unnecessary, makes them look like
suicidal or clueless idiots.

If you wish to distinguish two crucially important concepts, do not
name them with the same long words with multiple long adjectives then
needed. Use dissimilar short words
requiring zero adjectives.

I realize this a deep hard difficult concept in writing, but maybe
eventually if you work hard, you can figure it out.

[Jameson Quinn]

Hmm... having worked out an example of my system, I see that it's still too complex.

Let me try to simplify.

You rate the candidates in your riding, and can optionally delegate to one party or candidate outside your riding. If you don't, you implicitly delegate to your favorite inside the riding.

Highest score in each riding wins.

If a party has extra full quotas (that is, more quotas than the seats they won), they get one extra seat per quota. This goes to whichever has the most delegated votes.

Leftovers are transferred, starting with the lowest. The decision of what party to transfer them to is given to the non-winning candidates in proportion to how many delegated votes they got.

If votes are exhausted, quotas are adjusted. Parties win extra seats as they fill up quotas. As before, those seats go to the candidates with the most delegated votes.

This is basically the same as Warren's latest proposal, except that candidates cannot transfer votes outside the party until their party has the lowest number of excess delegated votes (less than 1 full quota). This does not affect independents at all.

[Jameson Quinn]

All three of us are highly atypical people. I would rather not trust the judgement of any of us three as to what words are best for communicating with more typical people, unless we all agreed. We should get a 4th opinion.

Anybody else reading this: what word(s) should we use to talk about the transferrable voting power for this set of proposals?

1. "juice"

2. "delegated votes"

3. "voting power"

4. "trust"

5. "ballots"

6. "delegated ballots"

7. "assets"

8. "vote assets"

9. "transferrable votes"

10. "votes"

11. "points"

Personally, I think option 1 makes us look uniquely bad, and any of the others would be fine if defined clearly and used consistently.

[Warren D Smith]

JQ: Balinski fair majority voting

WDS: Ah yes! -- Balinski's paper is here:
http://rangevoting.org/BalinskiFairMajGerry.pdf
And I had already discussed it here:
http://rangevoting.org/BalinskiPR.html

But is that really the same thing as my proposal?
Balinski never mentions the word "matching" or "bipartite" or "graph"
in his paper, even though my central problem is a bipartite graph
matching problem!

Balinski's central problem is a certain linear programming problem. Is his
LP problem, the same as my matching problem (by the way, my
graph problems definitely have LP reformulations)?
I think the answer is likely to be "yes" and the equivalence proof
will involve taking
the "dual form" of the linear program and then using logarithms!

If they are equivalent, then
it is a pity Balinski did not point that out
since it's quite helpful for comprehending what is going on and yields
a new optimality theorem that Balinski had not noticed, and new generalizations,
ditto.

(If they are not equivalent, also good to say so.)

The combined view of this from both the primal and dual views gives
you a lot more insight.
So in my opinion a new math paper should be written saying the above
and exploring those insights. Unless somebody already did.

But anyhow, returning to Canada, my old page had criticized Balinski's
system for
being susceptible to what I'd called the "monkeywrench effect."
That seems like a serious problem.
And Brams had raised the same issue JQ did about how the method
allegedly is unsellable. I have a hard time judging sellability, but
the monkeywrench effect
is a serious thing entirely apart from sellability, because it
converts third parties into
massive force-amplifying strategic-voting weapons capable of "targeted
killings."

It might be that this same sort of effect also is present in altered form,
in other riding-based PR schemes. So watch out.

[William Waugh]

"If you wish to distinguish two crucially important concepts, ..." What are those concepts again? If there were only one concept, I'd probably favor "points". But since there are two to distinguish, the terms should highlight the distinction.

[William Waugh]

The proposals mentioned to date seem remarkably complex and hard to understand.

[William Waugh]

In https://groups.google.com/d/msg/electionscience/g_JwRuUvwak/r1Nu3EfrBQAJ , Warren D. Smith posts another variant of his proposal. Let me highlight a couple of its rules:

4. Each party deserves a number of seats proportional to the total 
amount juice they own.

... each voter names her favorite party 

nationwide. 
2. Each time a voter names a party, that party gets 1 cup juice.

Scanning the proposal, one can verify that rule 2 is the only source of juice for a party. Rule 4 determines that juice is all-important for determining who gets elected to fill the T top-up seats, those being the seats to be filled after the representatives who won in their localities have been seated.

Consequently, much of the political power to be accorded by this system would come through a vote-for-one process. Therefore it seems reasonable to expect vote splitting, betrayal of favorites, and lesser evilism.

Why not permit the voters to approve or score more than one _somethings_, either proxy electors or parties or something, outside their own riding?

No one has responded to the proposal I put up, but I suppose the main objection to it is the complexity inherent in RRV, which I suggested for the top-up phase.

[Jameson Quinn]

OK, I'm reluctantly giving up on all the complexity. Here's my latest proposal:

You rate all the candidates in your riding, and can additionally name one candidate or party if you wish. (Independents count as if they were in a party of their own. Any party with M or fewer candidates is considered as a set of independent individuals, where M is somewhere from 1 to 3.)

You have one point to give. If you name a candidate, it goes to that candidate. If you name a party, it is divided between all the candidates in that party, in proportion to how many points they got directly. Otherwise, it is divided equally between the candidates you top-rated.

Top average score in each riding wins.

Points for each party are added up. If a party has enough points to deserve more full seats than they got from ridings, the non-winning candidates in that party with the highest points get those seats.

Points are used up for each seat awarded so far. These come from the winning candidates first. If more points are needed, they come proportionally from all other candidates in the same party. If that is still not enough points, that's just too bad.

At this point, all leftover points amount to less than one seat worth per party.

Whichever party has the fewest total points remaining is eliminated, and all candidates in that party may give their points to another party, or allow them to expire. If they give them to another party, the points are divided between the candidates in that party in proportion to how many points they got directly.

If at any moment, a party has over 1/(S+1) of the remaining points, where S is the number of remaining seats, the candidate in that party with the most points is elected, and that number of points are used up, as above.

Eliminations proceed in this fashion until all seats are awarded.

[Warren D Smith]

Returning to the "Balinski/Smith" system
(actually, I have not worked thru the proof my proposal was
really equivalent to Balinski's but, let's assume it is for the moment)
its two main flaws that we know about are

Flaw (a): As Brams & Quinn both pointed out, system allegedly unsellable
because it can make X win in a riding, and Y not,
even though Y is much more supported by that riding by
every popularity measure imaginable.

Related Flaw (b): As I'd pointed out, it can be used for "targeted killings"
via strategic voting. If the Republicans want to "kill" (un-elect)
a prominent Democrat who'd normally comfortably win his riding, they
just urge people in that district to vote for the CrazyLoons party.
Then that's the most-CrazyLoon
riding in the country, so that is where the one CrazyLoon seat will be, and
Prominent Dem is gone. Only a quite small number of CrazyLoon votes
can be needed to pull that off, it can artificially amplify the power
of the strategic voting tremendously.
If there then is a second Prominent Democrat in some other riding
they'd also like to
knock off, no problem, they just do it again this time using the
BeerLovers party
as the tool, and so on. So then all the Prominent Dems are gone
replaced by a ragtag randomly selected bunch of CrazyLoons,
BeerLovers, and on and on. Not good.

It seems to me both these flaws can be eliminated
simply by agreeing only to use Balinski/Smith for TOP UP purposes!

Eh? Then, Prominent Dem will win her riding just using the first stage
(e.g. score voting) system -- which just is not susceptible to
way-easy targeted killings.
Then the top up still will be susceptible, but we don't terribly care any more
and it is much harder to pre-identify "targets" for "killing" anyhow
since the top ups
are "nonlocal" MPs anyhow, and the PR depends on first-stage election
results not very predictable pre-election.. So therefore not a lot of
effort will be put into such targeted killings. And similarly since
they are nonlocal, nobody will terribly care that some riding greatly
preferred Y over X, but Y was elected --
because Y is viewed as elected not by his riding, but by whole country.

So if you bought that argument, then a pretty simple, entirely
vote-based, system
can be devised by just using score voting within ridings, then doing a top
up using Balinski/Smith to try to restore proportionality. (The
linear program is
not so simple, but that all doesn't need to be talked about in the
system definition.)
The seat counts for each party would be chosen after the local MPs were elected,
to best-approach PR, then run Balinski/Smith to fill the seats optimally.

As advantages it would be very simple to understand what it does,
and also would obey an optimality theorem.
I think it would enjoy some monotonicity properties, but you might
have a hard time
or be unable to prove nice "properties" mostly.
This scheme would have the disadvantage that it depends on explicit party IDs
and it's never been tried.

[Bruce Gilson]

On Sun, Oct 25, 2015 at 4:11 PM, Jameson Quinn <jameso...@gmail.com> wrote:

​[..

.

​]​

Both Clay and I have said we hate the word "juice". You say you hate the words "delegated votes". I don't understand your reasoning, but I think it has to do with the fact that my explanation was unclear in ways other than my choice of words for that concept.

​Can you and Warren perhaps come up with a word acceptable to both of you? Instead of "cup of juice" perhaps something like "vote credit"?

​

[Bruce Gilson]

On Sun, Oct 25, 2015 at 5:26 PM, Jameson Quinn <jameso...@gmail.com> wrote:

All three of us are highly atypical people. I would rather not trust the judgement of any of us three as to what words are best for communicating with more typical people, unless we all agreed. We should get a 4th opinion.

Anybody else reading this: what word(s) should we use to talk about the transferrable voting power for this set of proposals?

1. "juice"

2. "delegated votes"

3. "voting power"

4. "trust"

5. "ballots"

6. "delegated ballots"

7. "assets"

8. "vote assets"

9. "transferrable votes"

10. "votes"

11. "points"

Personally, I think option 1 makes us look uniquely bad, and any of the others would be fine if defined clearly and used consistently.

​Of the set, I prefer "points." ​

 

[Bruce Gilson]

On Sun, Oct 25, 2015 at 7:54 PM, Warren D Smith <warre...@gmail.com> wrote:

​[

...

​]​

a pretty simple, entirely
vote-based, system
can be devised by just using score voting within ridings, then doing a top
up using Balinski/Smith to try to restore proportionality.  (The
linear program is
not so simple, but that all doesn't need to be talked about in the
system definition.)
The seat counts for each party would be chosen after the local MPs were elected,
to best-approach PR, then run Balinski/Smith to fill the seats optimally.

As advantages it would be very simple to understand what it does,
and also would obey an optimality theorem.
I think it would enjoy some monotonicity properties, but you might
have a hard time
or be unable to prove nice "properties" mostly.
This scheme would have the disadvantage that it depends on explicit party IDs 

and it's never been tried.

​I rather like it. And as for "never been tried," It's not all that different from the ideas of score voting combined with MMP as done in Germany or New Zealand. So it uses ideas that are mostly familiar. 

[Clay Shentrup]

Juice is horrendous. Points is confusing, because you're already using Score Voting ballots.

I was originally thinking about something like "votes", or "awards", or "merits". But after some thought, I think "ballots" is probably best just because it's the most descriptive. It reminds me of picking names for variables in software. It just gives you an easier time of explaining the process.

But again, "juice" is like, put your face in your hands and shake it level awful.

[Clay Shentrup]

I dislike the explicit recognition of parties, and I suspect similar thinking partly accounts for those who prefer STV to MMP in Canada. I can't state any reasons that haven't already been aired here. I just think a system should generally leave as much as feasible to the "market". If people want to form parties, let them.

Thinking more about naming, maybe NP-MMP is decent = Non-partisan MMP. That's perhaps the most distinguishing feature about it. The scoring is important, but perhaps less of a selling point.

[Warren D Smith]

>> 1. "juice"
>> 2. "delegated votes"
>> 3. "voting power"
>> 4. "trust"
>> 5. "ballots"
>> 6. "delegated ballots"
>> 7. "assets"
>> 8. "vote assets"
>> 9. "transferrable votes"
>> 10. "votes"
>> 11. "points"

--Seriously? You consulted, maybe, a dictionary, and you were unable
to come up with a short evocative word clearly distinct from votes & scores?
After all that work you did writing all this down?

Well, you got "assets" for 1 out of 10.

[Jameson Quinn]

I would like to rewrite my latest proposal in simpler terms. I believe this proposal is nearly identical to one of the latest ideas proposed by Smith.

---

Begin with a list of candidates, each of which is either independent or a member of a party, and each of which is also running in a single riding. Any party with 2 or fewer candidates is considered as a set of independent individuals. (This does not harm them in any concrete way.)

You rate all the candidates in your riding, and can additionally "write in" one candidate or party if you wish. (Write-ins use a robust system to respect voter intent and avoid spoiled ballots insofar as possible.)

You have one delegated point to give. If you name a candidate, it goes to that candidate. If you name a party, it is divided between all the candidates in that party, in proportion to how many points they got directly. Otherwise, it is divided equally between the candidates you top-rated.

Top average score in each riding wins. (Scores are totally separate from the delegated points.)

For each seat awarded so far, N/(S+1) points are taken away, where capital N is the total number of points awarded (the number of voters who top-rated or wrote in at least one candidate), and capital S is the total number of seats to be awarded. These points are taken first from the winning candidate themself. If more points are needed, they are taken in equal absolute numbers from all other same-party candidates with points remaining. If all candidates in the party run out of points entirely, that party is considered to be eliminated, and no more points are taken for that seat.

If at any moment, the total points for all candidates of a given party is over n/(s+1), where n is the total number of points remaining, and s is the number of remaining seats, the candidate in that party with the most points is elected. In that case, n/(s+1) points are used up, as above.

Whichever party has the fewest total points remaining is eliminated, and all candidates in that party may give their points to another non-eliminated party, or allow them to become exhausted. If they give them, the points are divided between the candidates in the party in proportion to how many points they got directly. Points cannot be transferred between parties twice; they become exhausted when the second party they were held by is eliminated.

Awards and eliminations, as described in the last two paragraphs, proceed until all seats are awarded.

...

This has several advantages:

1. Almost all styles of voting which people are likely to try are at least reasonably close to strategically optimal.

2. Party gatekeepers have no role in voting, and the system in general gives little or no way for party members to reward/punish each other. This helps avoid the system devolving into corrupt "heavy whipping".

3. By the same token, the voters have the power to remove any unpopular candidate from office, even if that candidate has the support of their party colleagues.

4. Ridings are only a little bit larger than they are today, yet full proportionality can be achieved.

5. All voters have a local representative from their riding.

6. Voters can delegate to somebody they sympathize with ideologically, and the system is likely to use that vote for some candidate who is close to that ideology, insofar as possible.

7. Counting votes is simple and summable.

[Clay Shentrup]

On Sunday, October 25, 2015 at 8:09:14 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

--Seriously?  You consulted, maybe, a dictionary, and you were unable to come up with a short evocative word clearly distinct from votes & scores?
After all that work you did writing all this down?

"Ballots" is quite good. Descriptive and not silly sounding (like your brutally horrendous "juice").

[Warren D Smith]

"fuel" could perhaps be used instead of "juice."

Anyhow, this semantic stuff is hardly important especially at this stage.

[Warren D Smith]

http://rangevoting.org/CanadaSA1.html

has been rewritten to include discussion of criticisms and responses.
Some new.

[Rob Wilson]

Did your contacts say they are looking for some form proportional representation? Are they planning on trying to pass this through a referendum or are they going to ram it through? You might want to try to stick with something that a layman would be able to understand in a 30 second ad.

Also, does it have to be perfectly proportional? Perhaps you could just elect half proportionally irrespective of who wins in the single member ridings. That way you don't have to explain all the reweighing.

Are you sure that the LP would even be receptive to a system that promotes independent candidates over the party?  Perhaps you could suggest a system like Brazil that has an open list, but instead use asset voting within the party.

You don't want to propose something that the contacts will just dismiss after read the first paragraph. It is better to find the best system that has the potential to be implemented rather than the best system we can be conceive.

[Jameson Quinn]

As far as I can tell, the proposals are homing in on some pretty standard components:

-MMP

-Score voting locally

-Single vote, party proportional for national seats.

-Open list within parties

-Possibly delegation to deal with the fractional quotas and inter-party transfers.

Of those, score voting and delegation are the only "innovative" aspects, and both are in a context where even problematic systems wouldn't be too bad. In other words, I think that both score and delegation improve this system, but even if I happened to hate those mechanisms, I'd recognize this as an open-list MMP system, and as such pretty good.

[Warren D Smith]

Jameson, I posted one of your latest schemes as
http://rangevoting.org/CanadaSA2.html
after some slight editing for formatting & improved wording
etc which I hope you are ok with.

[Rob Wilson]

B. Party gatekeepers have no role in voting, and the system in general gives little or no way for party members to reward/punish each other. This helps avoid the system devolving into corrupt "heavy whipping."

Are you sure the Liberal Party would go for this?

[Warren D Smith]

Some data about Canada:
http://rangevoting.org/CanadaSeatsVotes.html

[Warren D Smith]

http://www.nationalnewswatch.com/2015/10/28/our-foreign-minister-at-sobeys-a-cautionary-tale-about-electoral-reform/#.VjDP42SrS2X

is a Canadian article basically saying that he's wary of PR schemes
that may detach
politicians from their constituents.

[Toby Pereira]

I've been on holiday so I'm a bit late to all this. One thing I'd note is that I find it strange that it's taken this this project to get people seriously thinking about national levels of proportionality with a non-party-list system. Jameson Quinn's PAL has been around for ages and while I've discussed it with him, I feel the general topic has been largely ignored.

I haven't read everything here, but from what I have read, I think this is very positive and the systems suggested seem to be largely in the ball park of how I think a fair system should be. I think I prefer Jameson's suggestion of pre-declared delegation rather than Warren's idea of candidates having to make decisions upon elimination. I think it is simpler and more transparent. I have also suggested before that a party candidate's delegated vote should pass directly to the rest of the party's candidates in equal measure, rather than having any layers of factions, both for simplicity and to stop "high ranking" members of the party from insisting that they are in everyone's top faction turning it into a more closed party list system. And it seems that Jameson has gone along these lines in contrast to PAL, even if it is just for reasons of simplicity rather than ideological reasons.

This has come up in the discussions but I'm not sure if it's been defined into the methods, but I don't think it's necessary to distinguish between zeros and blanks. I can see that it might be quite nice in principle to give the option to a voter to essentially give a candidate an average score, but it's an added complication, and realistically I don't think many voters would want to do this, so I think blank should equal zero. This needs to be no more complicated than necessary and this is not necessary.

I don't know if I've missed something, but I'm not sure what happens to the points of independent candidates in Jameson's system. Are they simply lost upon elimination? Or does an independent candidate submit a ranking list of other (independent) candidates/parties?

Also, when I considered mixed-member systems before, I thought it would be difficult to use score voting before moving onto what is essentially the STV stage. With plurality it's easy (even if it's a really undesirable system itself) because you vote for one and that's where your points go. The systems that have been described here are a bit of a compromise and not entirely ideal I would argue because we have this two-tier system of votes. If a candidate is elected with high scores from voters but not many top ranks, then for the second part of the voting they are considered to be over-represented, because the second part looks at the first part from a plurality point of view. But I think that goes against the whole ideology of what score voting stands for. Granted, it may be difficult to fix this in a simple way, but I think we need to acknowledge that this "kludge" exists as a problem of the systems. I might be mistaken but I don't think that this has been discussed.

A tentative potential solution would be to go fully proportional score for the second phase. Each candidate would have a predeclared score for all the other parties/candidates (could very well be zero for many but that doesn't matter). Each voter in addition to giving scores to the local candidates (and also potentially a top score to one write-in) indicates their favourite candidate as before. For the second phase, each voter takes on the scores of their favourite candidate for all candidates that they haven't explicitly given a score to themselves. So if a voter gives a score to all of their local candidates, then none of these scores will be affected. Only the scores of the non-local candidates would then be decided by their favourite candidate.

Once we have these full scores, the rest of the candidates are elected sequentially using a proportional score system. Which one to use is a fairly minor detail at present, but it can be discussed. I don't think it should automatically be RRV, however as I have strong reservations about it (discussed elsewhere).

[Clay Shentrup]

Another dissent:

http://rabble.ca/blogs/bloggers/skimber/2015/10/wake-harpers-defeat-lets-not-forget-electoral-reform#comment-2330667661

This system ignores the fact that proportionality for Canada must by province-by-province, since each province has a fixed number of MPs, following the general principle (with minor exceptions) that the number of MPs is proportionate to its population. It further ignore the fact that, within provinces, Northern Ontario voters will insist that their votes go to elect Northern Ontario MPs; and similarly for Eastern Ontario, eastern Quebec, the BC Interior, and many other regions. Scotland's model has regions of 16 MPs, Wales has 12.

It dances around the fact that, under the similar model used in Baden-Wurttemberg, some ridings have one MP, while others have as many as four. With such models, voters have the great benefit of competing MPs, one on the government benches, others in the opposition. But they should have a voice in the election of the additional MPs from the region.

This is an example of the fact that, when we have a model already designed by the Law Commission of Canada, which half the Liberal caucus supported in the House last Dec. 3, some people get excited about playing with models. The point is to implement PR, not to play with it.

[Jameson Quinn]

Toby asks:

 

I don't know if I've missed something, but I'm not sure what happens to the points of independent candidates in Jameson's system. Are they simply lost upon elimination? Or does an independent candidate submit a ranking list of other (independent) candidates/parties?

In all the systems I've proposed, when a party or independent candidate is eliminated, each eliminated candidate sends their delegated votes to the party they prefer. In the earlier versions, that was via predeclarations; in some of the later versions, that was an ad-hoc decision.

You raise a number of other interesting issues in your post but I don't have time at the moment to respond fully.

[Warren D Smith]

TP:

I have also suggested before that a party candidate's delegated vote
should pass directly to the rest of the party's candidates in equal
measure, rather than having any layers of factions, both for
simplicity and to stop "high ranking" members of the party from
insisting that they are in everyone's top faction turning it into a
more closed party list system.

WDS:
Sorry, that does not work, in the sense that it fails to yield PR.
You then would have situations where party candidates keep getting
eliminated, whereas with unequal distribution of "assets" mainly to
the "asset-poor" ones, they would stay alive and get elected. This can
even keep happening repeatedly.

[William Waugh]

What are the other proportional score systems?

On Wednesday, October 28, 2015 at 9:49:22 AM UTC-4, Toby Pereira wrote:

...

Once we have these full scores, the rest of the candidates are elected sequentially using a proportional score system. Which one to use is a fairly minor detail at present, but it can be discussed. I don't think it should automatically be RRV, however as I have strong reservations about it (discussed elsewhere).

  [ https://groups.google.com/d/msg/electionscience/g_JwRuUvwak/GmMLt4d0CAAJ ]