Electoral Reform: Neglected questions on mixed-member proportional representation

20 posts / 0 new
Last post
Jacob Richter
Electoral Reform: Neglected questions on mixed-member proportional representation

Does the district or geographical constituency component of mixed-member proportional representation always have to be based on plurality voting, and not preferential voting / ranked voting?

Does the district or geographical constituency component of mixed-member proportional representation always have to be single-member, and not multi-member?

Brian White created a Babble discussion in 2007 on combining the single transferrable vote system and MMP from an STV-preferred perspective:

http://archive.rabble.ca/babble/ultimatebb.cgi?ubb=get_topic&f=9&t=001594

Brian White wrote:
STV and MMP proponents were at each others throats in BC before during and after the referendum.(and it did nobody any good)
Perhaps a proposal like this might get most of them onside?

In his proposal, the added members would amount to only 25-30% of the total number of seats.

Coming from an MMP-preferred perspective, I would propose that no more than half of the total number of seats should be based on districts or geographical constituencies.  That lower limit lowers the risk of overhang.  That risk of overhang would be reduced further if all district / geographical constituency seats were to be elected on the basis of preferential voting / ranked voting, and if all but the most sparsely populated geographic constituencies were to be consolidated into multi-member ones.

I believe this is where such STV component under MMP would fulfill an invaluable "fair vote" role.

Doug Woodard

Jacob Richter wrote:

Does the district or geographical constituency component of mixed-member proportional representation always have to be based on plurality voting, and not preferential voting / ranked voting?

Does the district or geographical constituency component of mixed-member proportional representation always have to be single-member, and not multi-member?

Brian White created a Babble discussion in 2007 on combining the single transferrable vote system and MMP from an STV-preferred perspective:

http://archive.rabble.ca/babble/ultimatebb.cgi?ubb=get_topic&f=9&t=001594

Brian White wrote:
STV and MMP proponents were at each others throats in BC before during and after the referendum.(and it did nobody any good)
Perhaps a proposal like this might get most of them onside?

In his proposal, the added members would amount to only 25-30% of the total number of seats.

Coming from an MMP-preferred perspective, I would propose that no more than half of the total number of seats should be based on districts or geographical constituencies.  That lower limit lowers the risk of overhang.  That risk of overhang would be reduced further if all district / geographical constituency seats were to be elected on the basis of preferential voting / ranked voting, and if all but the most sparsely populated geographic constituencies were to be consolidated into multi-member ones.

I believe this is where such STV component under MMP would fulfill an invaluable "fair vote" role.

Jacob, in every MMP system known to me, the local constituency component has employed plurality voting in single-seat constituencies. However, I don't see any reason why that has to be the case, and I have spent a lot of time and effort informing myself about proportional representation.

You seem to have missed the point that CiaranQuinn made in 2007, that the larger the average number of seats per constituency in the local constituency component of an MMP system, the more proportional is the constituency component of the system (if a proportional system like PR-STV is used for the multi-seat constituencies), and the fewer list/top-up seats are needed to achieve an acceptable degree (or any given degree) of proportionality. The smaller the required fraction of list seats, the smaller can be the total number of seats. In past years, we have encountered some feeling that the total number of MPs should not be increased much, although that may have been related to the Mike Harris position that politicians are bad (or at least an extravagance) and we should have as few of them as possible.

Rein Taagepera and Matthew Shugart suggested in their book "Seats and Votes: The Effects and Determinants of Electoral Systems" ( Yale U. Press, 1989), that the optimum number of seats in a single chamber should be a compromise between ease of communication among MPs (favouring a low number of MPs), and ease of communication between MPs and citizens (favouring a high number of MPs). They concluded that the optimum number of MPs  would be the cube root of the number of politically active citizens, which they asumed to be the number who voted. We are already above this figure, which would be about 270 for us.

It seems that we have a strong feeling among the Canadian public that constituencies should not be too large for campaigning purposes, and that sparsely populated districts should be represented by their own and not by residents of adjacent or nearby cities. The upshot seems to be that people in sparsely populated areas will be unhappy with anything but single-seat constituencies, and that seats in single-seat constituencies might have to comprise about 40% of the seats in the constituency component of an MMP system.

The implication is that the fewer list seats we need, the smaller and more convenient the single-seat constituencies can be for any given total number of seats. This situation is favoured by having the constituency component of an MMP system comprise as many seats in multi-seat PR-STV constituencies as we conveniently can. This allows the constituency component to be somewhat proportional on its own and therefore to need fewer list seats to achieve whatever degree of proportionality we want. 

Another tradeoff is that if we are to have open lists (or even optional open lists) and the possibility of any voter choosing any list candidate of his/her favourite party in the list region, it is desirable that the ballot not be very large, therefore the number of list seats and the size of the list region should not be large. I think it is for this reason that Wilf Day has proposed regions with totals of 15 seats (constituency plus list), as much as possible.

As far as I recall, the MMP models for Canada ignore compensation for overhangs and acccept the damage to exact proportionality, for the sake of simplicity.

JKR

I think single-member ridings are very popular in Canada so STV is not going to be a viable option for the committee. However, single-member AV ridings will definitely be an option the committee might recommend for either a stand alone AV system or more likely a hybrid mixed-member system.

It will be very interesting to see what kind of process the committee sets up. Hopefully the committee will consult extensively with the public, electoral systems experts, and politicians from all political stripes. A process of compromise and consensus should help establish a "made-in-Canada" solution for electoral reform.

AntonyHodgson

If we simply paired adjacent ridings and used the ranked ballot to choose two candidates in each paired riding, the degree of disproportionality would be significantly reduced (by approximately 50%).  This enables us to have a much smaller top-up layer (probably less than 20% of all seats) to achieve the same degree of proportionality generated by MMP, which starts from a higher level of disproportionality due to its single seat districts and so requires more like 40+% of the seats to be in the top-up tier.  I'm calling this variant PR-MMP, for Paired Ranked Mixed Member Proportional.  It's similar in concept to Brian's idea in that it combines the STV idea of multiple representatives in the first tier (though only two) with the compensatory tier idea of Mixed Member Proportional systems.  Since Liberals like the ranked ballot and the NDP like MMP (and since Conservative Kim Campbell once proposed twinned ridings), there's something here that every party should like!

Sean in Ottawa

AntonyHodgson wrote:

If we simply paired adjacent ridings and used the ranked ballot to choose two candidates in each paired riding, the degree of disproportionality would be significantly reduced (by approximately 50%).  This enables us to have a much smaller top-up layer (probably less than 20% of all seats) to achieve the same degree of proportionality generated by MMP, which starts from a higher level of disproportionality due to its single seat districts and so requires more like 40+% of the seats to be in the top-up tier.  I'm calling this variant PR-MMP, for Paired Ranked Mixed Member Proportional.  It's similar in concept to Brian's idea in that it combines the STV idea of multiple representatives in the first tier (though only two) with the compensatory tier idea of Mixed Member Proportional systems.  Since Liberals like the ranked ballot and the NDP like MMP (and since Conservative Kim Campbell once proposed twinned ridings), there's something here that every party should like!

You would need more than a pair -- since this would average any mismatch to 50%. In parts of the country you could assemble 5 or so close ridings but in other parts the geographic area would be massive. This is an urban idea and makes less sense when you consider rural Canada.

that said it might not be as important as people may think that all MPs have consituencies that are geographic. List MPS with other connections that are less geographic might serve us well. I realize this would be two classes of MPs but one may not really be less than another -- just different. Having some MPs not have consituencies to attend to might free them up for other national and regional examinations.

AntonyHodgson

Sean in Ottawa wrote:

You would need more than a pair -- since this would average any mismatch to 50%. In parts of the country you could assemble 5 or so close ridings but in other parts the geographic area would be massive. This is an urban idea and makes less sense when you consider rural Canada.

I'm not sure what you mean here - with our current Single Member Plurality system (SMP, aka First Past the Post), about half the voters on average have voted for the winning candidate.  With paired ridings, typically about 80%+ will have.  This leaves only about 20% or less who need to find representation through other means - e.g., via the top-up component I described.  Of course, you're correct that overall proportionality would improve with larger combined ridings (4-5 seats per district) - this was the principle behind the Single Transferable Voting system that BC's Citizens' Assembly on Electoral Reform proposed.  Using a standard disproportionality index (the Loosemore Handby Index), our current SMP/FPTP system has a disproportionality of ~20% (and this doesn't properly reflect regional imbalance), while the paired ridings part of PR-MMP alone would reduce this to under 10%, and adding the top-up component would further reduce it to about 3%.  STV with district sizes of about 4 or so would typically have an index of ~5-6% - intermediate between Paired Ranked alone and PR-MMP.

Sean in Ottawa wrote:

that said it might not be as important as people may think that all MPs have consituencies that are geographic. List MPS with other connections that are less geographic might serve us well. I realize this would be two classes of MPs but one may not really be less than another -- just different. Having some MPs not have consituencies to attend to might free them up for other national and regional examinations.

I agree that it could be useful to have non-geographically-constrained MPs.  I can't think of a clear example of this anywhere else in the world, though New Zealand has parallel Maori electorates that each cover ~10 standard ridings, and Maoris can opt to vote in a Maori electorate rather than a general one.  It would be interesting to think about how this might be done in Canada, and how the individual voters could maintain appropriate ties to the MP they helped elect (and how that MP could be aware of where they've drawn their support from).

Doug Woodard

Jacob Richter wrote:

Doug Woodard wrote:
You seem to have missed the point that CiaranQuinn made in 2007, that the larger the average number of seats per constituency in the local constituency component of an MMP system, the more proportional is the constituency component of the system (if a proportional system like PR-STV is used for the multi-seat constituencies), and the fewer list/top-up seats are needed to achieve an acceptable degree (or any given degree) of proportionality.

I have argued for STV under MMP, not the reverse, because district or geographical constituencies in general still have one inherent problem: gerrymandering.

Keep the district or geographical constituencies to at most half the total legislative seats, and the risk of gerrymandering is lower.

Jacob, I think that the main risk of gerrymandering, and the main motivation for it (recall Sir John A.'s project of "hiving the Grits") is distortion of party representation. If constituency representation (whether single-seat or multi-seat) is nested withing an MMP list system whuch ensures overall party proportionality, the motivation for gerrymandering disappears. Not only that, but in multiseat PR-STV constituencies it is almost impossible anyways.

Jacob Richter

Doug Woodard wrote:
You seem to have missed the point that CiaranQuinn made in 2007, that the larger the average number of seats per constituency in the local constituency component of an MMP system, the more proportional is the constituency component of the system (if a proportional system like PR-STV is used for the multi-seat constituencies), and the fewer list/top-up seats are needed to achieve an acceptable degree (or any given degree) of proportionality.

I have argued for STV under MMP, not the reverse, because district or geographical constituencies in general still have one inherent problem: gerrymandering.

Keep the district or geographical constituencies to at most half the total legislative seats, and the risk of gerrymandering is lower.

EDIT: "The risk of gerrymandering" here refers specifically to the risk of gerrymandering to maximize overhang.

Jacob Richter

Doug, I apologize for not having been clearer above.  I added an edit to the post you responded to.

I agree with your first sentence completely.  My edit was a disagreement on your second sentence.  The greater the percentage of constituency seats embedded within MMP, the greater the specific risk of gerrymandering to maximize overhang, which goes back to your statement of distorting party representation.  That's why Brian White's framework is problematic.

To express disagreement with your last sentence, here are some specific cases of gerrymandering within standalone STV:

https://en.wikipedia.org/wiki/Gerrymandering#Australia
https://en.wikipedia.org/wiki/Gerrymandering#Ireland

To quote this paper: http://www.lse.ac.uk/government/research/resgroups/PSPE/pdf/2008conferen...

"To be practical STV requires relatively small districts that elect 5-6 members. ‘Such relatively small districts can be gerrymandered, which is highly undesirable in divided societies. List PR can easily be applied in much larger districts that are immune to gerrymandering’ (Lijphart 1991, 98-9)."

Doug Woodard

Jacob Richter wrote:

Doug, I apologize for not having been clearer above.  I added an edit to the post you responded to.

I agree with your first sentence completely.  My edit was a disagreement on your second sentence.  The greater the percentage of constituency seats embedded within MMP, the greater the specific risk of gerrymandering to maximize overhang, which goes back to your statement of distorting party representation.  That's why Brian White's framework is problematic.

To express disagreement with your last sentence, here are some specific cases of gerrymandering within standalone STV:

https://en.wikipedia.org/wiki/Gerrymandering#Australia
https://en.wikipedia.org/wiki/Gerrymandering#Ireland

To quote this paper: http://www.lse.ac.uk/government/research/resgroups/PSPE/pdf/2008conferen...

"To be practical STV requires relatively small districts that elect 5-6 members. ‘Such relatively small districts can be gerrymandered, which is highly undesirable in divided societies. List PR can easily be applied in much larger districts that are immune to gerrymandering’ (Lijphart 1991, 98-9)."

Jacob, point taken. I assumed that by gerrymanderring you were referring to manipulation of electoral boundaries and the types of voters contained within them, rather than the manipulation of district magnitude (DM, MPs per constituency) in multi-seat constituencoes.

Canada's present FPTP system is structurally extremely vulnerable to classical gerrymandering, but we have managed to keep it under reasonable control by keeping boundary changes out of the hands of politicians in recent decades. Also, despite Fianna Fail having driven the Irish DM down to 3.85, their PR-STV system manages to deliver quite decent overall party proportionality while avoiding the need for formal thresholds, and being friendly to independents and small parties like the Socialist Workers' Party - as long as they have a good appeal to transfer votes. 

I think that Paul Mitchell's LSE paper which you reference, makes a good case against Lijphart's pro-list position, and in favour of PR-STV.

I am heavily influenced by observation of what seems to be a strong sentiment in Canada in favour of single-seat constituencies at least for the low population density areas. I fear that in order to get any form or degree of proportional representation we will have to accomodate this feeling, quite outside the question of whether it is technically right or wrong. Also, I think that Canadians have a mistrust of lists - again this is significant for PR whether the popular sentiment is right or wrong.

I notice that keeping the constituency component of an MMP system down to 50% implies that the areas of the single-seat constituecnies have to be twice as big as now (assuming a constant number of MPs) which creates a conflict with the desire for single-seat constituencies (to ease campaigning and give "local representation") in the first place. If they have to be twice as big then why not make them 2-seat PR-STV constituencies in an overall PR-STV system?

If we are to have some kind of MMP system I like the idea of making the constituency component a mixture of AV/IRV single-seat constituencies, and PR-STV multi-seat constituencies, keeping the numbers of MPs elected by single-seat constituencies and from lists, both as low as conveniently possible.

We have to recognize that whether we get proportional representation or any approximation to it depends largely on the Liberals. and on the balance of power between two schools of thought within the Liberal party, the pro-AV/IRV side and the pro-PR side - and as well on how the Liberal strategists balance the prospect of ultimately having to cooperate with the NDP and/or the Greens in government, against the prospect of finally destroying and burying the CCRAP coalition, forcing it to either change or break up. We can offer the Liberals about as much ranked voting as they want - just not all in single-seat constituencies.

I am not overly worried about the risks of the various forms of gerrymandering in Canada, I think they can be dealt with by our now-traditional method of keeping the design of constituencies away from the politicians, and the inherent resistance (not absolute but very useful) of PR-STV to gerrymandering. The problem of gerrymandering seems to me comparable to the problem of party apparatchiks influencing the selection of list MPs. Both require attention but we can deal with them.

I am not much worried about the deleterious effect of some overhang on proportionality as long as we prevent it from acting together with thresholds. Perfect proportionality would expose us to the multiplication of small parties, uncertainty of coalitions, and general difficulty of negotiations within Parliament. I think it's a mistake to aim at perfection in general, and especially in party proportionality; there are always tradeoffs. I think it's essential to have some small parties if only because they are an essential part of the growth and decay of parties which I think we have to allow for, on the analogy of biological evolution. But I think that their systematic promotion on the classical Israeli model is probably undesirable. I think that it is highly desirable to have our system design make the success of politicians depend to a large degree on their contact with the voters, and with the voters' resultant opinions of them.  I suspect that some of the bad features of Israeli politics result from a lack of this kind of contact and judgement. I recall reading of a paper which concluded that Dutch voters (in their national-list system) were more likely to have met a member of the Dutch royal family than a member of parliament.

Also, I find formal thresholds unpalatable. I think it's one thing to say to people like Christian Heritage, Libertarians, Communists and Marxist-Leninists, "we're going to make things hard for you as parties, but as individuals you can still influence the result" and quite another thing to say in effect "we're going to ignore you completely." The latter I consider un-Canadian as well as politically unhealthy.

Jacob Richter

Doug Woodard wrote:
I am heavily influenced by observation of what seems to be a strong sentiment in Canada in favour of single-seat constituencies at least for the low population density areas. I fear that in order to get any form or degree of proportional representation we will have to accomodate this feeling, quite outside the question of whether it is technically right or wrong.

Indeed.  I don't think anybody here is advocating that the Yukon and the northernmost part of BC be combined into a single constituency.

Quote:
Jacob, point taken. I assumed that by gerrymanderring you were referring to manipulation of electoral boundaries and the types of voters contained within them, rather than the manipulation of district magnitude (DM, MPs per constituency) in multi-seat constituencoes.

Actually, at a very hypothetical level, there is still the very hypothetical possibility of classical gerrymandering under STV.  Where a single parliamentary party commands a legislative majority, that group can still engage in redistricting shenanigans: these are the electoral boundaries for so-and-so's four-to-six-seat constituency.

But thanks for pointing out district magnitude as a form a gerrymandering.

Quote:
Also, I think that Canadians have a mistrust of lists - again this is significant for PR whether the popular sentiment is right or wrong.

Supporters of lists, like me, simply have to do some more persuasion.

Quote:
I notice that keeping the constituency component of an MMP system down to 50% implies that the areas of the single-seat constituecnies have to be twice as big as now (assuming a constant number of MPs) which creates a conflict with the desire for single-seat constituencies (to ease campaigning and give "local representation") in the first place. If they have to be twice as big then why not make them 2-seat PR-STV constituencies in an overall PR-STV system?

Being twice as big as now is still nowhere near as "bad" as the shenanigans of the US House of Representatives or the Indian Lok Sahba.

Quote:
The problem of gerrymandering seems to me comparable to the problem of party apparatchiks influencing the selection of list MPs.

Having a pro-party perspective to civic participation, I don't see the basic principle of list / "parliamentary party" accountability to the party (within the framework of the party being accountable to its voters, of course) as a problem at all.

Quote:
as long as we prevent it from acting together with thresholds.

My "threshold" scenario is somewhere between what the Dutch have and what the Israelis used to have (now they do have a threshold), actually.  From the Dutch, if a parliamentary party wins a constituency seat, then it should be entitled to full list representation.  It's just that STV within MMP would make winning that constituency seat a lot easier... and thus actual left representation of social strata such as the precariat.

JKR

Quebec’s Liberals stand to benefit when electoral reform takes effect; Toronto Star; Chantal Hebert; May 28, 2018: https://www.thestar.com/amp/opinion/star-columnists/2018/05/18/quebecs-l...

Quote:
Earlier this month, three of the four parties in the National Assembly including the currently leading Coalition Avenir Québec signed an electoral-reform pact.

Should one of them win the Oct. 1 vote, its government would be honour-bound to introduce legislation within its first year in office to move Quebec to a mixed-proportional voting system. The pact is not legally binding but it does bear the signature of the party leaders.

...

In Quebec next fall, a vote for the CAQ, the Parti Québécois, Québec Solidaire or the Green party will be a vote for a mixed proportional system, to be put in place in time for the following election. The implementation of the Quebec pact is not conditional on a referendum.

...

In any event, in Quebec’s case, a CAQ or PQ government would not have a handy excuse to wiggle out of the pact as its signatories have agreed on a mixed-proportional system.

Mr. Magoo Mr. Magoo's picture

The Star wrote:
It is not the first-past-the-poll system that has kept the Liberals in power in Quebec for most of the past 15 years

Typo, or intentional?  Yanni or Laurel?

cco

If the PLQ would be the major beneficiary of MMP, as Hébert assumes, they've had 15 years (give or take a brief PQ interregnum) to introduce it, not to mention a big incentive after losing a wrong-winner election to the PQ. Not only would it let them count the supermajorities they rack up in the West Island, it'd give the PQ's rivals more air. Despite floating the idea halfheartedly a few times, they haven't done it. Why?

My hunch is that the PLQ is much happier competing under the current electoral system, where they can scare anglophones into staying in their right-wing big tent with the bogeyman of a referendum. Under PR, their internal coalition would run the risk of collapsing, with the Macpherson/Galganov types voting for some new version of the Equality Party, left-wing federalists bleeding to the NPDQ (or even QS, confident that no majority for sovereignty could be achieved), right-wing federalists bleeding to the CAQ or PCQ, a certain percentage bleeding to the Greens, and so forth. They'd lose their free hand on privatization, tax cuts, user fees, and tuition hikes. Eventually, they'd lose any remaining rationale for anyone to vote for them. Why pull the trigger on a change like that when "There's a referendum around the corner!" is the only campaign slogan they need?

Mr. Magoo Mr. Magoo's picture

Quote:
If the PLQ would be the major beneficiary of MMP, as Hébert assumes, they've had 15 years (give or take a brief PQ interregnum) to introduce it, not to mention a big incentive after losing a wrong-winner election to the PQ. Not only would it let them count the supermajorities they rack up in the West Island, it'd give the PQ's rivals more air. Despite floating the idea halfheartedly a few times, they haven't done it. Why?

I could be way off base here, and if you think so then please disregard this.  But it seemed to me that perhaps the claim is not that the Libs would be the lucky winners in a switch of electoral systems (because of the mechanics of the new system) so much as that they might do well if every other viable party supports that change of system (and promises to implement it) but the electorate doesn't care for that and votes Lib to avoid it.

JKR

Mr. Magoo wrote:

Quote:
If the PLQ would be the major beneficiary of MMP, as Hébert assumes, they've had 15 years (give or take a brief PQ interregnum) to introduce it, not to mention a big incentive after losing a wrong-winner election to the PQ. Not only would it let them count the supermajorities they rack up in the West Island, it'd give the PQ's rivals more air. Despite floating the idea halfheartedly a few times, they haven't done it. Why?

I could be way off base here, and if you think so then please disregard this.  But it seemed to me that perhaps the claim is not that the Libs would be the lucky winners in a switch of electoral systems (because of the mechanics of the new system) so much as that they might do well if every other viable party supports that change of system (and promises to implement it) but the electorate doesn't care for that and votes Lib to avoid it.

Where did she state that? It seemed clear to me that Hebert thinks the PLQ would benefit from PR as FPTP undervalues many of the votes they get in Anglophone constituencies where they get supermajorities.

From the article:

Quote:
More so than any other Quebec party, the Liberals pay a price for the high concentration of part of their vote. They enjoy massive support among non-francophone voters but that only translates into a small fraction of the province’s seats.

Take the current Quebec polls: province wide, the numbers suggest the Liberals are highly competitive. But in reality Couillard is in trouble because his party is trailing badly in the francophone ridings that will determine the outcome of the fall election.

It would be hard to craft a proportional voting system that would not be a gift that keeps on giving for the Quebec Liberals.

Jacob Richter

BC's three PR choices are Dual Member, MMP, and Rural-Urban.  I'm surprised Dual Member made it to the list, actually.

Rev Pesky

Jacob Richter wrote:

I'm surprised Dual Member made it to the list, actually.

What I can't understand is why the 180 other forms of PR didn't make it to the list. Why only those three?

Pogo Pogo's picture

 Do you think it was used on a Star Trek episode?  Given all the Trekkies in the NDP caucus.

Mr. Magoo Mr. Magoo's picture

Quote:
I'm surprised Dual Member made it to the list, actually.

Better that, though, than some other (presumably inferior) model making history, and the "Dual Member" club having to whine and bleat for the next hundred years that they were unfairly shut out by "Big PR".

I kind of agree with Rev. Pesky.  Put them ALL up the flagpole to see who salutes.  In terms of the long-game, it's better than any little clique feeling hard done by, and the rest of us forever having to accept their assertion that they're particular flavour would have won, hands down, if only the "fix" weren't in.

A three way race will have two losers.  A 180-way race will have 179 losers.  Losing fairly must be better than feeling like a disqualified winner.