Ontario, like other provinces and Canada as a whole, has a winner-take-all voting system. If you vote for the most popular candidate in your riding, you elect a member of provincial parliament and win representation. If you vote for another candidate, you elect no one and are unrepresented. In the last Ontario election, won by Dalton

McGuinty’s Liberals in 2003, almost a million Progressive Conservative voters went unrepresented, as did more than half a million NDP voters.

When Ontarians go to the polls again next October, it may be the last time they elect a provincial parliament under this system. As they elect their MPPs, voters will also be deciding on a proposal emerging from the deliberations of the province’s Citizens’ Assembly on Electoral Reform. The Citizens’ Assembly is about to begin public hearings as it works its way toward a recommendation to put on the referendum ballot. While we won’t know until next May what that recommendation will be, it will very likely be some form of proportional representation (PR).

Seventy-five countries, including most major democracies, use PR systems. The goal is to maximize the number of citizens who can help elect representatives they want: to make every vote count. To meet this objective, Fair Vote Canada calls for “a voting system … designed to achieve the following objectives: proportional representation; fair representation for women, and for minorities and Aboriginals; accountable government; real voter choice; and geographic representation. Provinces and regions must have effective and accountable representation in parliaments and governments, reflecting real geogra-phic communities.”

There are a variety of proportional voting models, including List PR and the Single Transferable Vote (STV). All of them satisfy fair voting principles better than the current system. However, the form of PR that I find combines Fair Vote Canada’s objectives best for Ontario is Mixed-Member Proportional (MMP). In this article I set out a made-to measure model of MMP, tailored to Ontario’s particular needs.

MMP from New Zealand to Quebec

MMP is not a new idea. It was first created by British authorities in West Germany in 1946. Merging British direct elections with European proportional systems, it balanced personal accountability with party account-ability. MMP is now used in Germany, New Zealand, Scotland, Wales and elsewhere. Quebec is about to present the latest version of its MMP proposal. New Brunswick was preparing to hold a referendum on an MMP model, which Premier Bernard Lord supported, until Lord’s government was defeated in the September 18 election. The Law Commission of Canada recommended a federal version in 2004. No two MMP models are exactly alike, since each one responds to a different set of objectives and geographical, political and cultural considerations.

Scotland’s simple model has the most appeal here, but the Scots weren’t faced with the challenges represented by Ontario’s vast size. Because the option presented here seeks to meet Ontario’s particular needs, its workings are more complex than Scotland’s, although the choices facing the voter and the actual procedure of voting are still relatively simple.

MMP is a “mixed compensatory” form of proportional representation: a mixture of local representatives and compensatory or “top-up” members allocated to compensate for the local results. While supporting the principle of proportionality, most Ontarians are wary of any change which will replace single-district representatives with legislators elected from lists. MMP is designed to allay these concerns. It uses lists like other PR systems, but at the same time it retains single-member districts. Every voter has both a local representative and regional representatives chosen from the lists (in New Zealand, the entire country is one region).

On voting day you get two ballots. You cast a local ballot for your local member, as we do today, and you also cast a second ballot for your choice of party. On that second ballot is a list of the party’s candidates for your region. One of the ways in which different MMP systems vary from one another is in their rules for allocating seats from the lists. Further on, I will set out a proposal for an allocation formula tailored to Ontario’s needs based on the objectives set out by Fair Vote Canada. But the underlying principle is simple: parties that win few local seats in relation to their popular vote get the bulk of the list seats, while those with more local winners than their popular vote would warrant get few if any.

Already, citizens’ groups in two provinces have designed MMP models. In Quebec, a group of eight representative citizens, after participating in months of public hearings, reported this April. It called for an MMP model with 60 per cent local and 40 per cent regional representatives along with a provincewide “top-up” calculation to give full proportionality. There would be a National Assembly of 129 members from 16 regions, ranging from 18 local plus 12 regional MNAs from Montreal down to three MNAs (two local, one regional) from Gaspésie. A party would need to reach a provincewide threshold of 5 per cent to get seats.

The British Columbia Citizens’ Assembly designed an MMP model in 2004 before deciding to endorse a Single Transferable Vote proposal instead. (This proposal won a majority in a referendum, yet fell just short of the 60 per cent threshold imposed by the B.C. government.) The model I am proposing is quite like the Quebec citizens’ model, but with open lists as in the B.C. Citizens’ Assembly’s model (the Quebec citizens favoured a closed-list model, meaning that party nominations can put minorities in a good list spot, and voters cannot “break the slate” by changing the party’s list ranking). It also goes further in taking into account the needs of people outside metropolitan areas.

Of course, Ontario will also be considering other systems besides MMP, including the Single Transferable Vote. Ireland’s STV model was one of the electoral systems studied by a select committee of the Ontario legislature in 2005, in preparation for the work of the Citizens’ Assembly. In Ireland, people vote by ranking candidates in multi-member constituencies. The committee’s only witness who liked STV recommended it for the two thirds of Ontario’s ridings that are in multiriding urban areas. However, she was far less enthusiastic about it for the other one third of the ridings: “The problem with STV in rural ridings is that to have a multimember district you would have territories that are so huge it would be impossible, unfeasible, to even campaign in them … In the north, it would just be enormous.”

For Ontario, MMP is a better fit. But overly large ridings can be a defect of MMP as well. This is less of a problem in metropolitan regions, where many voters don’t even know what riding they live in and where regional MPPs would be as accountable as local ones. But one third of Ontarians do not live in metropolitan areas. Some of their communities could lose their representatives in the legislature under MMP, and this could be unacceptable in the sparsely populated north, where many ridings are already vast. We will see how this and other challenges can be dealt with as we set out how the proposed MMP model would work and the advantages it would convey.

Designing an MMP model

Citizens face five main design decisions for an MMP model for Ontario:

  • the size of the legislatur
  • the ratio of compensatory regional seats to local seats
  • the number of regions
  • province wide proportionality, or proportionality within unlinked regions
  • closed or open lists

The five decisions are connected, and there are tradeoffs to be made between them.

The size of the legislature

Under the current law, Ontario will have 107 MPPs after the next election. However, in 1999 Premier Mike Harris’s government cut the size of the legislature by 27 MPPs. We restore those MPPs and have 135 seats, of which 83 would be local and 52 (39 per cent of the total) would be regional “top-up” seats. This makes it possible to keep the north’s 11 local ridings, as well as the 19 less urban local ridings in the south. These regions would have relatively few compensatory regional members: three (one per region) in the north, and five (in four regions) in the nonmetropolitan south.

The metropolitan regions of Ontario will be divided into 77 ridings for the 2007 election. In the model, instead of those 77 there would be 53 local ridings, which would be on average 45 per cent larger than the current ridings. Those ridings would be grouped into eight electoral regions with 44 regional MPPs. These regionally accountable MPPs would compensate for the larger local ridings.

16 map1The province would thus be divided into 15 regions: eight metropolitan regions (five in the Golden Horseshoe stretching from Durham Region east of Toronto to Waterloo in the west and then southeast to Niagara, plus Ottawa–Cornwall, London–St. Thomas and Windsor–Chatham) and seven others in northern Ontario and the less urban parts of the south (see map on p. 140). There would be provincewide proportionality: the regional “top-up” seats would be calculated provincially to correct the imbalances of the local results, and then allocated regionally. This lets us use different ratios of regional and local MPPs in different regions, while still counting each vote equally provincewide.

The ratio of regional to local seats

In existing MMP systems, the ratio of top-up seats to local seats varies from 33 per cent in Wales to 50 per cent in Germany. In recent Ontario elections, a third of the seats would have had to be allocated to top-up members to achieve proportionality. This figure will be higher when more voters vote their first choice, as is likely to happen in a PR system, so a legislature with about 40 per cent top-up members, as proposed for Quebec, gives a safety margin.

The number of regions and the nature of proportionality

The question of how many regions Ontario should have is closely tied to whether proportionality should be provincewide or within unlinked regions. A look at Scotland, where there are eight unlinked regions, shows how this works.

Suppose a party wins all nine of the local seats in a 16-seat region with only 45 per cent of the vote. Nine seats represent more than the proportion it should have, and therefore the seven top-up seats will not be enough to give the other parties their fair share. Thus, a party that wins one eighth of the vote but no local seats should get two of the seven list seats, but may end up with only one. Since this happens in half the regions in Scotland, the most popular party by far, Labour, ends up receiving a “winners’ bonus.” Last time it got 38.8 per cent of the seats with only 29.3 per cent of the votes, a result more typical of “winner-take-all” systems. These problems can be mitigated with fewer regions giving less chance for regional variation, as in the four regions proposed for New Brunswick, but four vast regions in Ontario would make regional MPPs less accountable. As well, for a self-contained region to have a member for a party with 4 per cent of the vote, it would need 25 members.

The solution is to borrow the Scandinavians’ method of linking regions to the overall vote. The Scandinavian system is not MMP – there are no local members of Parliament. However, Sweden and Norway have smallish locally accountable electoral regions; in Sweden they average 12 MPs each. To create nationwide proportionality, they add a “second tier” consisting of about 11 per cent of the MPs. There are no national party lists: the seats for a given party go to regional best runners-up (the candidates next on the list in the regions where that party was closest to winning another seat). In this way some regions get extra MPs, affecting regional balance. That’s not something Scandinavians pay much attention to, but in Ontario regional imbalances would be seen as unfair.

We need to find a way to ensure provincewide proportionality while making sure that the number of seats in each region is fixed. It is possible to do so and I will show how. But the model will have to be more complex than Scotland’s.

Let me cast the problem in concrete rather than abstract terms. In a region of 200,000 voters with five MPPs, say 80,000 (40 per cent) vote Liberal: they elect two MPPs. Say 54,000 (27 per cent) vote NDP: they deserve 1.35 MPPs, and get at least one. Say 50,000 (25 per cent) vote PC: they deserve 1.25 MPPs, and get at least one. And say the other 16,000 (8 per cent) vote for the Green: they deserve 0.4 MPPs. Who gets the fifth MPP?

The Green says, “We’re the largest minority. By the highest remainder rule we do.” The New Democrat says, “Hold on – with two MPPs we would have 27,000 voters for each MPP. You Greens have only 16,000 votes, so we do.” An STV fan says, “Neither of you has a good claim. Let’s decide by seeing who is the second choice of the 10,000 surplus Conservative voters.” “No way,” answer the Greens and the NDP in unison. “Why should we let Conservatives decide between two candidates of the left?”

There is another option. We can say, “None of you has a clean claim on this seat. Let’s see which party is underrepresented in the province as a whole and use the seat to correct the imbalance.”

In my proposed model, things would not be quite so simple in practice. Since a number of regions would have only one regional seat, the provincial results would be applied only in the larger regions. But for Ontario as a whole, allocation of the regional seats would compensate for the imbalances in the local results. The method for allocating regional seats is shown in Table 4.

Closed or open lists

Once the regional seats have been allocated among the parties, there still remains the question of who is actually going to fill those seats. In a “closed list” system, the parties decide. Each party presents a slate of candidates, and the voter’s only option is to decide which slate to vote for. If a party is entitled to five regional seats in a given region, the top five candidates on that party’s list become MPPs. Most MMP models have closed lists.

German law requires that all local nominations and regional list nominations be democratic, either by vote of all party members or by delegates elected by all party members. Ontario should have a similar provision. Still, if regional MPPs never face the voters directly, there is no mechanism to stop parties from filling their lists with obedient unknowns. Therefore, some PR models have “open lists” in which all representatives face the voters, who can influence not only the choice of party but also the choice of particular candidates for that party.

I am proposing an open list MMP model, but designing such a model may be the biggest challenge for the Ontario Citizens’ Assembly.

At the federal level, the Law Commission of Canada’s 2004 study recommended adopting Sweden’s new “flexible” list system, where voters have the option of voting for the party list or casting a “personal vote” for an individual on the list. To “break the slate,” an individual must get personal votes equalling at least 8 per cent of the party vote in the constituency. Any list candidate who exceeds this threshold is ranked by number of votes received, ahead of party-ranked candidates. In the Commission’s example, the “Red Party” wins four list seats in a constituency. Personal votes for the fifth candidate on the Reds’ list equal 10 per cent of the Reds’ party vote, while personal votes for the seventh candidate equal 8 per cent. Therefore, both of these candidates break the slate, and they get two of the seats. The other two go to the party’s two highest-ranked candidates.

One of the difficulties in designing a flexible list system of this kind comes in deciding what weight each personal vote should be given in relation to party votes. The 8 per cent rule makes sense in Sweden, where the 29 regions have, on average, 12 MPs each (8 per cent being roughly one twelfth). Sweden’s new model helped a couple of women break the slate in 2002. However, my proposed model for Ontario uses “natural regions” of sizes varying from four MPPs to 37. This variation makes the model more responsive to local conditions: voters can choose known names from the list. However, it also means that the 8 per cent threshold does not fit.

To weight single preferences fairly in relation to list votes, my version of the Swedish system is to give each single preference as many points as the number of list MPPs the party has won in that region. Thus, if a party wins five list seats in a region, each of that party’s voters in effect has five preferences. A vote for the slate then counts as one point for each of the top five candidates. A single preference for a particular candidate amounts to five points for that candidate.

More considerations to think about
Women and minorities

When party members in a region put together their list, they can nominate whom they like – proven winners, new faces, candidates who are also running locally or additional candidates. That is the party’s face as it competes for government. In a competitive world, Ontario can’t afford not to use all its human resources. Parties will want to respect the diversity of the electorate. This tends to lead, all over the world, to more women being nominated.

In the last Ontario election, only 21 per cent of the MPPs elected were women. When parties nominate candidates one at a time, they usually pick men. Lists give women more chance of an equal voice, and minorities more chance of election. With an open list system, voters still choose who gets elected from the list. Once women get past the gatekeepers of the single-seat ridings, won’t we vote for them? Ninety per cent of Canadians want to see more women elect-ed.

Longer lists likely help women and minorities more. Lists can, and should, have as many names as the total to be elected from the region. Local candidates usually run on the list too. Anyone elected locally drops off the list. The list needs a couple of spares in case of death or resignation of a regional MPP, when the seat goes to the next on the list. In a region of five MPPs, where a party hopes to elect as many as three, it needs a list of at least five, big enough for the position of women to be visible. In my proposed model, of 52 regional MPPs, 39 are from regions with lists of seven or more, and 34 from lists of ten or more.

Thresholds to win seats

In most PR systems, a party needs to exceed a threshold to gain representation in the legislature. What should that threshold be? Scandinavia and many others use 4 per cent. Some use 3 per cent. Germany and New Zealand’s 5 per cent is about the highest. Israel has just come up to 2 per cent; Palestine uses this too.

In favour of low thresholds, it is argued that every vote should count equally, as far as possible. High thresholds keep newer views out of the Legislature. Lower thresholds favour free expression of minority opinion. On the other hand, a threshold should be high enough for the legislature to be workable, and not unduly subject to extremists or vanity parties. With a multitude of miniparties, with only one or two members each, the legislature can become chaotic. Hitler took advantage of Wei-mar Germany’s instability. Canada extends federal election financing to any party getting more than 2 per cent of the vote, but 2 per cent may be too low for a threshold.

A 3 per cent threshold would have kept the Greens out of the Ontario legislature in 2003: they got 2.8 per cent. However, PR would likely have empowered Green Party supporters enough to bring them up to 4 per cent. In the 2006 federal election, the Greens got 4.7 per cent in Ontario.

If challenged in court under the Charter of Rights, the threshold must be shown to be a reasonable limit. If it was set by an independent Citizens’ Assembly, the courts might show more deference than if politicians with a conflict of interest had set it. A threshold of 4 per cent, it seems to me, would be about right: high enough to allow for a stable legislature but not so high as to stifle minority views.


In an MMP system, independents can still run for local seats. Most MMP models do not let independents run for the top-up regional seats, but there is no reason why they couldn’t. Scotland, for example, lets independents run for regional seats too. A winning independent would remove that seat from the provincial calculation and regional assignment of party seats.

The twin party problem

Politicians who want to manipulate the MMP system may try to run “twin parties”: one party label for their local candidates, and a second for their regional list candidates. If the two are recognized as separate parties, the list winners, rather than compensating for local disproportionalities, are added to the local winners. This gives large parties a big bonus. Parties are usually reluctant to risk alienating already cynical voters, yet this has happened twice already. The best solution is an elections act that gives the chief electoral officer strong powers to rule that twin parties are in reality a single party.

Accountability of regional members

Another challenge in a proportional representation system is ensuring that members elected from party lists remain accountable to a particular group of electors. This MMP system contains built-in incentives to ensure that this happens. If a party wins lots of local seats, it gets few top-up members, so that being on the list is no guarantee of election. As a result, regional members hope to run locally as well as regionally next time. This means that they compete with local members to serve constituents, and have their own constituency offices. Bad for the politicians, perhaps, but good for the voters. n

Wilfred Day has been a lawyer in Port Hope, Ontario, for 35 years. He is a member of the National Council of Fair Vote Canada. He has been called an electoral mechanic, which he takes as a compliment.