By Jameson Quinn, November 2015. [Last time I checked JQ was a grad student at Harvard. Slight editing by WD Smith which hopefully did not significantly alter anything JQ had to say. WDS is not necessarily endorsing any of JQ's claims or proposals here; he is merely re-posting them. Warning: This proposal by JQ is different from his previous (Oct. 2015) proposal.]
Posted here (/CanadaSA4.html) just just to serve as a reference point for further discussion.
As far as I can tell, the proposals for Canada being discussed now are homing in on some pretty standard components:
Of those, score voting and vote-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.
Prologue: Recently, Warren Smith started a thread about proposals for a PR system for Canada. In much back-and-forth in that thread, there were several proposals. The common thread was MMP, open list, with a good single-winner system at the district (riding) stage, and some asset-inspired system at the regional (province) stage.
I now want to propose something simpler. I have two ideas. The first uses a Score/MCA/GMJ hybrid at the riding level and an asset/open-list hybrid provincially; the second is the same, but at the riding level it uses... plurality. I realize that plurality is bad and we want to end it. But as a component of a PR system, it's actually pretty sensible.
So... here are the rules:
We shall regard voters as "female" and candidates as "male" just to make wording clearer.
0. Start with lists of ridings, candidates, parties. For each riding, make a ballot type which lists the local candidates.
1. [Optional] Candidates predeclare preference orders over the parties they consider allies.
2. Voting.
3. The ballots are tallied. The physical counting and reporting happens locally, of course, but for the purposes of determining winners, each riding's ballot type are tallied together, no matter where they were voted.
4. Find a winner for each riding using that riding's ballot type. If any candidates are "preferred" by a majority (i.e. over 50% of the voters for that ballot type), the most-preferred wins a seat; otherwise, the one with the highest total of "preferred" plus one-half "approved" wins.
4a (optional) Also figure out who would have won in each riding using only local ballots. Those candidates will go to the front of the line for provincial seats.
5. Tally up "points" for each candidate. Each ballot gives one point, divided equally between preferred candidates.
5a. (optional) Candidates with fewer than a certain fixed number of points are ineligible to win seats. (This rule works like Germany's 5% party threshold, to prevent small parties from accumulating seats based on a small percentage in each riding, and thus encourage a certain amount of pre-negotiated coalitions. Unlike the German rule, however, this does not lead to wasted votes, and can be circumvented by parties with good organization to concentrate their votes in one riding, or with natural strongholds in certain ridings. Another advantage of this rule is that, by discouraging excessive party splitting, it effectively increases the size of the plurality required for winning riding seats, and thus reduces the "proportional overhang" of larger parties. So in effect, this rule actually helps medium-sized parties, but it would be easy to sell to the large parties.)
6. Find total points for each party. Find the number of full Droop quotas each party has, and thus the extra seats they have earned above what they won at the riding level. Assign these seats to the non-elected party candidates with the most points. If you were using optional step 4a, those candidates get seats first.
7. Find the leftover points for each party. Eliminate the party with the smallest leftover, and redistribute those points to other parties, according to the preferences of the candidates from the eliminated party (proportionally by points). When any party reaches a full Droop quota (calculated by non-exhausted votes and remaining seats), they get a seat, assigned as in step 6.
8. Repeat step 7 until done.
The plurality version of this system is obvious. It's exactly the same, except that there is no "approved", only "preferred". Note that this is technically approval, as it allows a voter to have multiple "approved" candidates. But strategically, it reduces to plurality, because the strongest strategy for the extra seats is to prefer only one.
I don't have time right now to point out all the advantages and disadvantages of this proposal, but I think it's pretty good. Yes, people would complain about "carpetbagger" votes (or whatever term they'd have for that in Canada), but with rule 4a, that would not actually be a problem. And it's really quite radically simple for a proportional system, especially from the perspective of voters.
The other innovation in my proposal is that voters can effectively opt to vote in a riding of their choice, though their local riding is the default. The proportional part of the system works to help avoid "hostile takeovers". This is, as far as I can tell, the simplest way to avoid overlong ballots while still leaving good options for people who like party X but not the local party-X candidate.