Critical history of Canadian voting-system reform attempts 2000-2015

Warren D. Smith, Nov 2015

We will describe all the most prominent attempts to reform (and suggestions for reforming) Canada's electoral system during 2000-2015. Up to 2015, all failed to be enacted. In many cases (we shall see) this failure was deserved, in the sense that the proposals clearly are easily improvable and/or contain major flaws – or at least severe risks those are flaws. Which we'll point out. For our own proposals, which largely or entirely avoid those flaws, and seem better and/or simpler, see http://RangeVoting.org/CanadaOverview.html.

Law Commission of Canada report (pdf) of 2004

A serious proposal was produced for Canada (at the federal level) by the Law Commission in 2004, but their report was ignored and the commission disbanded.

Even though the report was 232 pages, it could not bring itself to mention approval and score voting, and in general most or all modern ideas (post-1950) are ignored.

On p.175 it finally makes its "Recommendations" which are (edited by me for conciseness):

  1. Add an element of proportionality to Canada's electoral system.
  2. Specifically, adopt an MMP (mixed member proportional) system.
  3. The MMP system should be based on giving voters TWO votes: one for a constituency representative and one for a party list.
  4. 2/3 of parliament should be FPTP single-winner-in-riding seats. The remaining 1/3 should be party lists (in each province or territory, not nationwide lists). [I've omitted some further minor details.]
  5. Parliament should adopt a flexible list system that provides voters with the option of either endorsing the party "slate/ticket," or of indicating a preference for a candidate within the list.
  6. Parliament should require political parties to develop initiatives and policies to promote equal representation of women... and increase racial minorities... and youth... and aboriginal peoples... and submit reports describing how they addressed those issues.
  7. A Parliamentary committee should subsequently review the parties' reports.
  8. A party should be eligible for compensatory provincial list seats only if it presents candidates for election in at least one-third of the constituencies in the relevant province.

I think they already blew it on item 3 (see also item 8) because independent MPs are inherently treated massively unfairly. The whole issue of independent MPs is totally ignored by their report. Item 4 also stinks since FPTP is simply a bad way to choose single winners. Item 5 is also suspect because plurality-style vote-totals are probably a bad way to compute a list-ordering – average scores using ratings-style ballots (or even approval-style ballots) seem far superior. Also, one could have made everything simpler for voters via PR systems only needing one ballot, not two.

British Columbia electoral reform referenda of 2005 and 2009

The "Citizens' Assembly on Electoral Reform" wanted to replace the existing "first past the post" (FPTP) in single-member ridings system, with a single transferable vote system (BC-STV). BC-STV won by 57.7% in 2005 (1.7 million voters), but this referendum was not binding unless the support exceeded 60%. Then in 2009 a binding referendum was held (1.6 million voters), whereupon BC-STV lost by 61-39 landslide, hence the FPTP system was retained. (Both these referenda put together total 51.31% against BC-STV.)

It is noteworthy that the proposed BC-STV system was not described on the referendum itself, probably because its proposers felt it was too complicated to describe. The exact wording of the referendum was:

Which electoral system should British Columbia use to elect members to the provincial Legislative Assembly?

The BC-STV system also was not described in the STV proponents' video (https://www.youtube.com/watch?v=y-4_yuK-K-k), although it did provide an extremely vague sketch. The BC-STV system also was not described in this 20-page official Final Report CAER put out, although that also contained a vague sketch. Apparently that was because even a 20-page report still was not enough to handle actually defining the system. For the definition, readers of the "final report" were referred to the "Technical Report" claimed to be "available at public libraries, universities and colleges throughout the province."

Oh.

Let us imagine the reaction of a voter, especially one who hadn't been paying much attention, upon seeing this choice. "What the hell does BC-STV even mean?" is her first question. (The referendum question could not even be bothered to unpack the acronym!) Perhaps a copy of the 20-page Final Report is available for perusal at her polling place, but, while very glossy and full of beautiful color pictures of smiling people and abstract art, it simply does not contain a system definition. (It does admit its absence and refer people to local universities, though. Thank you ever so much.) That's it, game over, says any voter even remotely like me. There is no way in hell I am voting for this proposal if they cannot even define it in their 20-page report. I have absolutely no idea how the "Citizens' Assembly on Electoral Reform" possibly imagined they could possibly achieve referendum success in this manner!

Suggestion: the definition of the system should fit on a single page, distributed at polling places, otherwise forget about it.

But I finally tracked down the actual BC-STV definition stated in that technical report, and here it is! (There's also this PDF containing a flow chart on page 2, and this "detailed example.") The whole proposal would basically convert BC into a pretty close copy of Ireland.

The old FPTP system involved 79 ridings throughout BC in 2005 (but 85 in 2009), each with population about 50000. In the BC-STV proposal ridings would be combined to become about 5 times the population, i.e. now 250000 each. Each riding now would elect not one, but rather 5, representatives. (Actually, numbers between 2 and 7 were permitted, but apparently 5 was typical.) Each ballot would be a rank-ordering of the candidates, but voters would be allowed to rank-order only some (indeed, even only one!), not all, of the candidates, if they desired (the rest being automatically ranked co-equal last). If, say, 7 candidates ran for each of the 5 seats (note there were 7.3 candidates per seat in the Canada-wide 1993 federal parliamentary elections) then each voter could rank all 35 candidates in rank-order – a difficult and error-prone task – or fewer. This is very distortionary because if you don't know anything about a candidate hence intentionally choose not to rank them, then you probably did not think they were the worst candidate ever, co-equal with Hitler and Stalin. Actually, that probably would be the largest possible distortion of your opinions.

The video claims that the BC-STV system is "proportional" and will give a party 40% of the seats "as nearly as possible" if they have 40% of the votes. That's PR – proportional representation. (This, by the way, is not at all obvious from the system definition.) However, since each riding elects 5 winners, the only allowed seat-fractions are 0, 20, 40, 60, 80 and 100%, hence errors in proportionality of up to ±10% of the total number of seats in parliament could happen. That would be far more than enough to completely zero the Green Party, which got only 6.78% even in its best-ever vote-percentage result Canada-wide in 2008. So under the BC-STV "proportional" system, if adopted Canada-wide, we would expect the Greens to get zero seats, even though, proportionally speaking, they deserved 21 seats in 2008.

Oops.

Further problems: The BC-STV system suffers many "pathologies," aka "paradoxical self-contradictions." For example, it can exhibit nonmonotonicity where a voter by raising a candidate on her rank ordering, can make him lose, and by lowering the candidate can make him win. It can be more-strategic to not vote at all, than to cast an honest rank-ordering vote. BC-STV can refuse to elect a candidate who would defeat any of his rivals by a large margin in a head-to-head 2-choice election. It also can elect somebody and regard them as the "best" candidate while if all ballot rank-orderings were reversed (so we were now trying to figure out the worst, not the best candidate!) it still would elect that same candidate and still contend he was the "best"! Also, it is simply false to contend (as some did) that STV eliminates the "wasted vote" problem – there can still be "spoiler" candidates who lose but, by running, prevent X from winning (if they dropped out, X would win). A vote for such a spoiler sure sounds "wasted" to me. And many of these pathologies occur frequently; they are not rare.

Also, BC-STV could only be counted centrally. It would be impossible to count it in precincts.

Also, if voters tried the naive but highly appealing (about 85% of Australian voters use this strategy) strategy of (1) determine the two parties most likely to win, (2) rank them top and bottom (or reverse), (3) rank the other parties in between somehow, then we are guaranteed to get extremely disproportional results yielding massive 2-party domination – which incidentally is exactly the situation in Malta, one of the most important users of a system of the BC-STV kind.

On the bright side, I like the fact BC-STV produces PR without need of party labels. Christy Clark (Liberal MLA who became BC Premier in 2011) in this 2009 video explains why she was opposed to BC-STV in 2005 but in 2009 supported it (thanks to Ron Campbell for pointing it out). More pro-STV youtube videos by Dennis Pilon: 1, 2, 3, 4, 5, 6, 7, 8, 9.

Why did BC-STV lose? Three reporters at TheTyee explain. Bill Tieleman, president of the "No STV" campaign, explains. Professor Dennis Pilon explains [Canad. Polit. Sci. Review 4,2-3 (June 2010) 73-89]. Canadian govt report discussing the 2005 referendum. Essentially, BC-STV was unappealing, complicated, would have made ridings so large that many people would have felt unrepresented, and the Liberal Party's support for it dropped perhaps due to their realization they have tended to benefit from disproportional representation.

An Angus Reid poll on 15 May 2005 found that 66% of BC voters claimed to know "nothing" or "very little" about BC-STV (the referendum was held 17 May), A Nordic Research poll (May 31 to 1 June 2005) found that 53% of those who voted against BC-STV did so because they did not feel knowledgeable.

The 2009 BC referendum would have required 60% overall supermajority plus ≥50% approval in at least 60% of the province's electoral districts in order to succeed (same rules in Ontario and PEI referenda we discuss below). Dennis Pilon branded those rules as

The Liberals clearly stacked the deck against any change occurring by establishing a super-majority threshold 60%... for STV... despite the fact that there existed no compelling legal or historical precedents for doing so [D.Pilon: The politics of voting, Emond Montgomery Publications 2007 pp. 103-4]. While some Canadian scholars argue that these measures were not deliberately designed to ensure defeat [Carty, Blais and Fournier 2008; 144, 157], more comparative accounts tend to characterize these decisions as a strategy of "process manipulation" by the politicians [Alan Renwick: The politics of electoral reform, Cambridge Univ. Press 2010 p.77].

Ontario electoral reform referendum of 2007

The proposal was to stop electing Ontario's provincial parliament via First Past the Post (FPTP) single winner ridings. The "Citizens' Assembly on Electoral Reform" instead proposed a simple mixed member proportional representation (MMP) system mildly resembling Germany's Bundestag. Each voter would cast two votes:

  1. one for a local candidate, and
  2. one for a political party.

The local member would be elected according to the usual FPTP mechanism. (Which is peculiar because the proponents of this reform all agreed FPTP was a bad system!) However, there would also be a number of "list seats." The second ballot would be used to determine what share of total seats (local + list) should be proportionately assigned to each party to ensure proportionality of the overall result. The proposed Legislature would have 129 seats consisting of 90 local members (70% of the Legislature) plus 39 list members (30%). List members would be chosen in order from a ranked list of candidates proposed by each party before the election.

What if you are an independent (not affiliated with a party!) candidate? Oh sorry: you are intentionally, by design of the law of the land, shafted. That's what they call "reform."

The parties' "deserved numbers of seats" would be computed from their vote shares (from the 2nd part of the ballots only) via the Hamilton-Vinton largest-remainder rounding-off method (formerly used in the USA for "apportionment," but later abandoned). This rounding off method is known to suffer from the "Alabama paradox," which is why the USA abandoned it. Hamilton-Vinton isn't horrible but the consensus opinion backed up by computer simulations is that it appears to be worse than Sainte-Laguë's method.

The referendum question exact wording (English only, I'm omitting the parallel French wording) was:

Which electoral system should Ontario use to elect members to the provincial legislature?

The MMP proposal was defeated by 63-37 landslide margin.

The "closed list" system removes individual accountability, assuring election of those near the top of their party's pre-decided list. To try to counter that criticism: "Before the election, parties will be required to submit their lists, as well as the details of the process they used to nominate their list candidates, to Elections Ontario – a non-partisan body. Elections Ontario will publish this information widely. Voters will be able to assess whether parties created their lists in a fair and transparent way."

Also, the proposed system still retains all the well-known flaws of FPTP in electing its local MPs. They could have, for example, switched that part to approval voting, an easy and clear improvement, but no. The whole point of a "local MP" in systems of this ilk, is to be the best choice in the opinion of his riding. So we should use a method which is good at producing single-choices, not a bad such system like FPTP, which is what Ontario chose in a clear design mistake. (Plus, this change as a beneficial side effect would probably make the local MPs more proportional, thus reducing the number of needed MMP top-up seats.)

Prince Edward Island 2005, New Brunswick 2008, and Quebec electoral reform referenda

These were three other MMP proposals, each apparently quite similar to Ontario 2007. (PEI's proposal involved d'Hondt's rounding-off method instead of Ontario's Hamilton-Vinton, and a 5% threshold for small parties to access list-seats.) PEI's lost by even larger 64-36 landslide. New Brunswick planned to hold a referendum on this early in May 2008, but didn't. Finally, Quebec's effort apparently died even before proposing a referendum date. PEI's question wording: "Should Prince Edward Island change to the Mixed Member Proportional System as presented by the Commission of PEI's Electoral Future?"

Why didn't it pass? Jeannie Lea [Canad. Parliamentary Review (Spring 2006) 4-8] claims that

  1. the committee designing the proposal complained unsuccessfully that they did not have enough time,
  2. the plebiscite rules demanded a 60% supermajority for passage,
  3. it was a standalone plebiscite, not held on the same day as any other election,
  4. the "no" side spread distrust/fear about the reform,
  5. the main political parties did not support the reform.
Note (e) was unsurprising in the sense that the main parties automatically tended with FPTP to be over-represented in parliament, so any PR reform was contrary to their interests.

If we ask why the similar Ontario 2007 reform did not pass, then the reasons are probably similar. William Kuchapski [Why Electoral Reform Failed in Ontario? Manitoba Law Journal 36,2 (2013) 137-156] claims they were "partisan self-interest, electorate apathy, and a lack of information" about the proposal provided to the public.

There also is a 60-page long report on the Ontario failure by L.LeDuc et al [later condensed to Lawrence LeDuc: The Failure of Electoral Reform Proposals in Canada, Political Science 61,2 (Dec. 2009) 21-40], which claimed "The press was hostile, and the public information campaign inadequate... Three quarters of those polled in the first week of October [the referendum date was 10 October] indicated that they had heard 'only a little' or 'nothing at all' about the issue... The Citizens' Assembly was an unknown entity for most of the electorate... There was relatively little information to be had, and only the most pro-active voters were likely to find it... Modern campaigns cost money, and the lack of any serious funding of the campaign for MMP undercut its effectiveness and deprived voters of the information that they needed to properly evaluate the proposed reform... There was no real need for an organized NO effort; the doubts raised by skeptical journalists more or less did the job... The more complex the issue, the greater the difficulty of putting it across to voters in a short campaign..."

Although the pro-PR forces probably were motivated by the BC-STV failure then to push simpler MMP systems... strangely enough, all the MMP defeats were greater-landslide defeats than the BC-STV defeat.

Year-2015 proposals by FairVote Canada as videos (each 6-8 minutes long)

A. "MMP proposal" https://www.youtube.com/watch?v=D3guVBhKmDc (Video narrated by Dennis Pilon, associate professor of political science at York University and one of the founders of FairVote Canada.) 297 local MPs, plus 131 regional MPs in 27 top-up regions. Voters cast vote for both local rep, and for a regional rep, using open lists. Each ballot has two parts:

  1. vote-for-1 local MP.
  2. vote-for-1-party (or for 1 candidate, which counts as a vote for his party) for regional MP.

Part 2 is considerably larger, since contains many more candidates. In the "sample ballot" they gave at 3:12 for "Ottawa South" riding in "Eastern Ontario" region, there were 5 local candidates, and 46 regional candidates, from 5 parties. None were independent. It's pretty absurd to make voters choose among 46 candidates. I do not believe most voters will be able to keep track of that many. Amazingly considering the huge size of their ballots, voters are forced to provide only an incredibly small amount of information with the "name one then shut up" style voting. The whole issue of independent MPs was simply not even mentioned during the entire video. At 4:12 they claim no strategic voting, and no vote-splitting, can exist with their method. Those claims both are simply false. First of all, in part 1 of the ballot, since it is just single-winner plurality voting for the local MP, there can easily be vote-splitting between alike candidates, and there can easily be strategic decisions not to "waste your vote" by voting for Ralph Nader (to allude to the USA 2000 presidential election) thus inspiring voters to dishonestly vote for a candidate who is not their true favorite. Second, in part 2 of the ballot, suppose your favorite party is going to get X seats, and your vote for that party clearly is not going to be enough to raise that to X+1 seats. But, if you lie in your vote by voting for a different party (which you also like, although less) that would be enough to give them 1 extra seat. In such a circumstance, you indeed would be an idiot to tell the truth in your vote, and strategic voting is wise.

They then "simulate" the election results based on year-2011 data finding at 5:10 that the 10 local ridings in Eastern Ontario would elect

Conserv=6, Lib=3, NDP=1, Green=0 seats (total=10 local);
then the top-ups would elect
Conserv=2, Lib=1, NDP=2, Green=1 seats (total=6 top-up); 
for in all
Conserv=8, Lib=4, NDP=3, Green=1 (total 16 seats).
The video does not describe the precise way they determine the party shares from the votes, instead merely making the vague claim that 37% of the votes will yield 37% of the seats. Well, 37% of which votes? (Remember, there are two kinds of votes, since your ballot has two parts.) They do not say. And obviously, in their 16-seat example they just gave, 37% of the seats would be 5.9 seats. That is not possible. So there will be some distortion caused by integer roundoff at the very least, causing each party's seat-count to be off by up to ±3% of the total #seats in parliament.

B. "Jenkins-inspired MMP proposal" https://www.youtube.com/watch?v=dOyHLwYq6Nk (Video again narrated by Dennis Pilon, associate professor of political science at York University.) Rank-order ballot for local MPs. At 1:42 "on its own, single member preference voting [IRV] does not create [proportionality]." Hence (1:55) their plan is to add "top-ups in regions" to convert IRV in ridings into a "semi-proportional system." 203 local MPs, plus 135 regional MPs, in 42 top-up regions. At 3:00: "The small size of the top-up regions would mean the parliament would not be fully proportional" by which I presume they mean that the obstacle is rounding-off to (necessarily small) integers. (Or, since the video calls this only a "semiproportional" system, for some unexplained reason, perhaps they have some additional obstacle in mind.) Obviously that same roundoff obstacle occurs in their proposal A too, but they there had pretended there was no obstacle! Well, I think what they are trying to say is, with proposal A there were 27 top-up regions, but with B there are 42, so therefore with A they can get more accurate proportionality within each (since larger) region. At 4:09 they again consider the typical(?) "Ottawa South" riding in "Eastern Ontario" region, now there were 5 local candidates running for 1 local seat (and there would be 5 ridings, hence 5 local seats, in this region), plus 24 candidates vying for 3 regional top-up seats. A voter rank-orders the 5 local candidates (and then one is chosen via the IRV process) and also names a single one of the 24 candidates (or simply names one party). Then by a mathemagical process they do not describe, somehow the top-ups are elected in a way that attempts to restore proportionality (and they also do not explain exactly what proportionality means – e.g. it is a share of what votes?). In their example, this time the Green party gets zero seats, as contrasted with proposal A where the Greens got 1 seat, from Eastern Ontario. Again, independent MPs are nowhere mentioned in the entire video. This proposal inherently makes it impossible to count ballots in precincts and forces centralized counting.

Non-definition: It is astonishing to me that in neither video A or B, did FairVote Canada actually define the PR voting process they had in mind! They used vague happy-sounding words, but there is no precise definition. Ever.

Here's a hint. If your system is so hard for you to define that you cannot do it in an 8-minute professionally produced video, then either it is too hard to define, or you are incompetent. Or both.

It's hard for me to issue precise criticisms, given that, I repeat, no precise definition of any voting systems has been given in either FairVote video A or B.

I will say this, though. The reliance on plurality-style "name one candidate then shut up" ballots in (A), and on rank-order ballots in (B1), both are design mistakes. Approval-style, and rating-style, ballots respectively are superior. This superiority is in terms of (1) lower ballot-error rates, (2) faster voter speed, (3) inherently provides more information, (4) to some extent avoids "Arrow theorem" and related inherent impossibilities, (5) ratings ballots allow voters who wish to express ignorance, to do so without introducing distortion, (6) "dumber" voting machines can be used.

Also, I again suspect that if voters employ the naive-exaggeration strategy of ranking the top two parties artificially first and last, then both these systems will force 2-party domination and massive disproportionality.

C. "Stephane Dion's P3 proposal" https://www.youtube.com/watch?v=wyLeClCrfgQ; I did not actually watch this video because Dion had provided written versions of his proposal as https://stephanedion.liberal.ca/en/articles-en/p3-voting-system-canada (short version) and a more detailed longer – 15 pages – version [Which Voting System is Best for Canada? Federal News 3,4 (April 2012) 1-15].

1. Ridings each would elect 5 MPs, not one (and the ridings each would be 5 times larger than now). ("Or perhaps four or three when a low population density warrants it. There would be some exceptions – such as the territories which, for practical reasons, would remain one-member districts. But the standard would be five-member districts.")

2. At the ballot box, voters would rank-order parties in order of their preferences. (Also allowed: ranking some but not all, in order. Including: just picking one.) Second, voters would select their preferred candidate among those put forward by the party they select as their top preference. Let's say that a voter chooses the Liberal party as his or her top reference; this voter then would choose one candidate among the five liberals.

3. Here's how to count. First, the voters' top-preferences would be counted. If one or more parties failed to obtain enough first choices to win at least one of the five seats (namely, I presume, at least 1/6 of the votes?), then the party that got the smallest number would be eliminated and its voters' second choices would be transferred to the remaining parties. The successive choices of the voters for eliminated parties would be allocated until all of the parties still in the running obtain a least one seat. The precise number of seats each of the remaining parties then got would then somehow be determined by some rounding-off procedure that Dion does not actually specify(!) but I suppose could be chosen from among the usual suspects.

4. Then, the voters' choices as to their preferred candidate among those attached to their preferred party are counted. If a party obtains two seats, that party's two candidates who received the highest number of votes would win those two seats.

Flaws in P3:

  1. Suppose the Liberal party runs 5 candidates A,B,C,D,E in some riding. A is unanimously agreed by all voters in that riding, to be the best person in the universe. So they all vote Liberal+A. But now the liberals win all 5 seats, which is bad because B,C,D,E were all corrupt bastards, and honest Tories would have been superior. Point is, B,C,D,E "rode A's coattails" into office. P3 creates inter-candidate correlations, via parties, which is not necessarily good.
        The voters could instead try to avoid this fate by voting Tory, but then A can't get elected.
        Either way, it stinks. Or, the voters could vote dishonestly. That might work better, but if so, the fact this is required, is not good.
        Think that scenario is unlikely? Well, one way I think it actually might realistically arise in the form where it wasn't the liberals, it was the Greens, and A was some world-renowned environmentalist, but Greens B,C,D,E were nobodies likely to be disasters as MPs...
  2. P3 would elect zero Greens even in Canada's 2008 election featuring 6.78% Green votes (its best-ever vote-percentage).
  3. Tory voters are not allowed to express their preferences among the Liberal candidates.
  4. What about independent (party unaffiliated) candidates? Blatantly unfair to them.
  5. If voters employ the naive-exaggeration strategy of ranking the top two parties artificially first and last, then P3 will force 2-party domination and massive disproportionality.

I think P3 is an interesting "ad hoc" design, hopefully working pretty well for the very particular numbers that Dion expects. I like Dion's transferable-vote-among-parties idea as a way to assure proportionality (albeit if only 5 MPs are elected, then large proportionality errors are inevitable including necessarily "zeroing out the Greens"), but: it forces the voters to supply rank-orderings, which is a problematic ballot-type; and it forces the system to be counted centrally, making precinct-based-counting impossible; and it causes P3 to massively fail NESD.

Conclusion

All the proposals above have major flaws. None of the three year-2015 proposals has actually been fully specified. None of them employ modern ideas (post-1950) except for Dion's P3 system, which seems quite original.


Return to main page