Friday, September 27, 2013

Stéphane Dion's moderately proportional P3 model: how well would it work?


Stéphane Dion, the former Liberal Leader, spoke at the launch of Fair Vote Canada’s 2015 campaign, in September 2013.

He said “we are numerous to think, in the Liberal Party, that we need some moderate proportional representation. Early next year we will have a convention and we are already looking for new ideas. Justin Trudeau strongly believes that Canadians need better choices when they enter the voting booth, and is committed to ensuring the party’s grass-roots have a real voice in the party. So, the debate is going on in our party.
“The preferential ballot does not solve the main problem I identified, which is that the current electoral system exaggerates artificially the regional support of our parties and makes the country look more antagonistic than it really is. For this reason I will continue to argue that moderate PR should be added to the preferential ballot as a solution, and I will make my point well-known in my own party. My aim is that all political parties should commit to truly look at ways to improve our electoral system.”

Indeed, the Liberal Party is no longer committed to the preferential ballot alone. Their 2014 convention voted overwhelmingly for a priority resolution from their federal MPs caucus that stated “THAT immediately after the next election, an all-Party process be instituted, involving expert assistance and citizen participation, to report to Parliament within 12 months with recommendations for electoral reforms including, without limitation, a preferential ballot and/or a form of proportional representation, to represent Canadians more fairly and serve Canada better.”

Dion’s reasons why the preferential ballot alone does not solve the main problems are very persuasive, in his original article and on his current website.

Dion’s P-3 model: Moderately proportional five-MP regions
Instead of electing a local MP, you would elect five MPs from a region of about 500,000 people, proportionate to the votes in that region.

With five MPs in a region, your party needs 16.7 per cent of the votes to elect an MP. If it gets fewer, your vote is transferred to your second choice of party.

This is different from the mixed systems recommended by the Law Commission of Canada and the Jenkins Commission in the UK, and used in Scotland and Wales. In mixed systems you still have a local MP accountable only to the local community. You also have competing regional MPs elected from a “top-up” region to make the results proportional to the votes. This system guarantees a local competitor among the field of regional MPs, who each have a mandate from a segment of the regional community.

The Law Commission of Canada noted that mixed systems are thought to combine the “best of both worlds”: the accountability strengths of single-member constituencies, along with the demographic representativeness and fairness of proportional representation systems.

More people would vote, and vote differently

As Prof. Dennis Pilon says: "Now keep in mind that, when you change the voting system, you also change the incentives that affect the kinds of decisions that voters might make. For instance, we know that, when every vote counts, voters won't have to worry about splitting the vote, or casting a strategic vote. Thus, we should expect that support for different parties might change."

And when every vote counts, turnout will be higher -- perhaps 7% higher. So, when voters have more choice, the results will be far more representative of our diverse population and their diverse views. Who can say what would be the result of real democratic elections?

One thing we know for sure: it is extremely unlikely that voters would vote exactly as they did in 2011.
Meanwhile, I’ve done projections on the votes cast in 2011.

How would P3 work in practice?
So how would Stéphane Dion’s “moderate proportional” P-3 model work? Let’s look at the likely results for the new 338 MP-House, using the votes cast in 2011.

Elections Canada has transposed those votes to the new boundaries. With the new 30 MPs, the winner-take-all results for the 338 MPs would be 188 Conservative, 109 NDP, 36 Liberal, 4 Bloc, and 1 Green.

In my simulation of Dion’s model it works, moderately proportionally, as intended. It would not greatly increase the number of parties represented in Parliament.

Dion’s P3 reflects votes much more accurately than winner-take all: 43 fewer Conservative MPs, nineteen more Liberal MPs, nine more New Democrats, and 15 more Bloc MPs, but no more Greens.

NDP voters and Liberal voters elect 173 MPs, a potential majority coalition with no need for the Greens. The Conservatives and Liberals would have 200 MPs, an alternative potential majority coalition.

But the Greens will likely do better, getting more votes. The Greens could easily double their vote, and start winning maybe six more seats under Dion’s P3, see below.

Regional diversity
None of the major parties is shut out of any region, Dion’s main objective.
Western Liberal voters elect seven MPs, including one from Calgary now shut out of their caucus, one from Surrey, and one more from Winnipeg. That’s better than their present four, although short of the 12 Western MPs full proportionality would have given them.
Conservative voters elect one MP in Montreal, one in Quebec City, two in Victoria—Nanaimo, one in Halifax, one in Downtown Toronto, one more in Scarborough, one more in Hamilton, and one more in PEI. Better in Quebec than their present five, but well short of the 13 Quebec MPs full proportionality would give them. Better in the cities.
Quebec Liberal voters elect two more MPs: one in Longueuil, one in the Outaouais. That’s nine instead of seven, although two short of the 11 MPs full proportionality would give them.
Personal votes
Dion’s model lets voters whose first choice is a certain party vote for their favourite of that party’s regional candidates.
This will help women. Parties would have an incentive to diversify their offering of candidates in each district, more representative of women and cultural diversity. Polls show 94% of women want to elect more women MPs, and so do 86% of men.

Since I’m projecting from the 2011 votes, I’ll start with the 2011 candidates. For example, Greater Toronto Area Liberal voters could have elected Martha Hall Findlay, Yasmin Ratansi, Maria Minna, Karen Mock, Bonnie Crombie, and Ruby Dhalla. Maybe Anita Vandenbeld in Ottawa, Marie Bountrogianni in Hamilton, and Karen Redman in Kitchener. For Ontario New Democrats, Myrna Clark in Barrie, Dr. Wendy Wilson in Parry Sound, Susan Galvao in Cambridge, Nadine Hawkins in Markham, and Michelle Bilek in Mississauga.

Preferential: EKOS poll of second choices
Dion’s model is both proportional and preferential. For example, if your Green Party candidate doesn’t get enough votes for election in your region, your vote counts for your second choice of party.

Dion’s model lets voters rank the parties. My simulation from the votes cast in 2011 uses a poll by EKOS just before the election, asking voters what their second choice would be.
Should a coalition government prove necessary, P3 would prepare parties to make it a stable and coherent government. By seeking out the transfer of subsequent voting preferences from their respective voters, the parties would be encouraged to highlight similarities in their platforms and objectives, and would be better prepared to govern together.

Competing MPs
Competition among the five MPs in a single district would provide Canadians with better territorial representation. Voters would be able to deal with the MP of their choice in their district, instead of having only one as is the case now.

No more strategic voting?
Because Dion’s model has a preferential ballot, it will greatly reduce the need for strategic voting.

However, it will not end it totally. Take the case of a Green voter who knows the Greens will not win one of the five seats in her region, the NDP will, and the top two NDP candidates are a woman environmentalist and a male union leader. She really wants the woman to win. Dion’s model allows her to cast a personal vote, but only for a candidate of the party that was her first choice. She will vote NDP in order to cast her personal vote for the woman she wants to see elected. Since 32 regions would have elected no Liberal in 2011, some Liberal voters could have faced the same issue.

How five-MP regions work
Because Dion’s model has no local MPs, he wants the regional MPs to come from regions small enough that MPs “would all have a home base,” and the degree of proportionality is ”moderate.”
Dion’s standard is five-MP regions, so my simulation uses five-MP regions whenever possible. It ends up with 51 five-MP regions that elect 76% of our MPs. Geography and simple mathematics dictate some four-MP regions, and in remote regions, even smaller. So the rest are in 14 four-MP regions, seven three-MP regions, one remote two-MP region (Abitibi-Nord-du-Quebec), and four isolated single seats (the Territories and Labrador).

Good for stronger parties
Of Green Party voters, 33% stated the NDP as their second choice, 17% the Liberals, and 16% the Conservatives. This helps the NDP.

Dion’s P3 works well for 2011 NDP voters as a whole. They elect 11 more MPs from across Ontario, six more from Alberta, five more from BC, four from Saskatchewan, two more from Manitoba, and two more from New Brunswick. This more than makes up for electing 15 fewer Quebec MPs.
It also works pretty well for Liberals in Ontario. In my simulation, Ontario Liberal voters would elect 15 more MPs, 28 MPs, only three short of the 31 MPs full proportionality would give them. The small regions give Ontario Conservative voters a bonus (compared to full proportionality) of five MPs and NDP voters three, at the cost of the Green party’s potential five MPs and the Liberals’ shortfall of three.

No local MPs 

The first and biggest problem with Dion’s model is, in large parts of Canada, the lack of a local MP accountable to your community.

Canada’s metropolitan areas will have no problem with Dion’s P3. Five-MP regions, with 500,000 people each, would work well for large communities: Toronto, Montréal, Calgary, Ottawa, Edmonton, Mississauga, Winnipeg, Vancouver, Brampton, Hamilton, and Québec City. But 68% of Canadians live in municipalities too small to have five MPs. It could work in the rest of Greater Vancouver Region, York Region, Durham Region, Halton Region, Waterloo Region, Laval, Longueuil, and the ridings of four more MPs in Montreal suburbs. Four-MP regions could work in Halifax and Niagara Region.

But across Canada, of the 338 new MPs, only 168 are from those communities. That still leaves out 49% of Canadians. By the 2011 census, 60.0% of Canadians live in population centres bigger than 100,000 people. That leaves 40.0% who don't.

It’s worst outside Quebec and Ontario.

In Atlantic Canada, 28 of the 32 MPs would be from regions much larger than their communities.

Six of Manitoba’s 14 MPs, all 14 of Saskatchewan’s, 14 of Alberta’s 34, and 19 of BC’s 42 MPs – 51% of Western MPs – would be from regions far larger than their communities.

In Ontario, it leaves Ontario’s smaller communities with no MP accountable to their community. Since they now each elect one of the 43 Ontario MPs outside those larger metropolitan areas, this model will be a very tough sell there. To them, having no MP except from a 500,000-strong region just doesn’t sound like democracy. Even medium-sized centres as large as Sarnia and Chatham could find themselves represented only by MPs from large urban centres. Smaller centres and rural areas would surely feel left out.

That’s why the Law Commission recommended a model inspired by Scotland’s. With “the best of both worlds” you have both a local MP and a group of regional MPs from your larger region. The party you voted for may have elected only a regional MP this time, but you are guaranteed he or she must compete with your local MP in the four years before the next election.

A total of 170 MPs don't fit Dion's regions. Of these, six MPs are in isolated single seats and the two-seater. The other 164 of the 332 MPs to be elected from multi-MP regions – 49% of them -- are lumped into regions much larger than their communities.

In Quebec the 2013 Electoral Boundaries Commission Report notes “the great attachment many expressed toward their RCMs” (Regional County Municipalities, like Ontario’s regions and counties). But Quebec has 86 RCMs and 14 more equivalents. Outside Montreal, Laval, Longueuil and Quebec City, they are all too small for a five-MP district. The average RCM has a population of 40,000. You can partner two or three of them in a district of 100,000.You can even partner four of them in a larger district of 150,000. But in a region of 500,000 people, they are all swamped.

In some suburbs it might work. Longueuil is the size for four MPs; perhaps the 74,416 people of RCM Marguerite-D’Youville will feel represented even though added in to a five-MP Longueuil. Laval and Montreal have 22 MPs, four five-MP regions and two left over: maybe the 307,637 people of the three ridings in RCM Les Moulins, RCM Thérèse-De Blainville and RCM Deux-Montagnes will feel at home with 200,000 people in Laval.

But this still leaves 43 of Quebec’s MPs from regions far larger than their communities.

Less than full proportionality
The second problem with Dion’s P3 model is lack of proportionality.
If we use province-wide totals with perfect proportionality (not counting Quebec Greens who got less than 3% of the vote), voters would have elected 139 Conservatives, 104 New Democrats, 65 Liberals, 19 Bloc, and 11 Greens. NDP, Liberal and Green voters would elect 180 MPs, a potential majority coalition, needing the Greens for the majority. For an alternative potential majority coalition, the Conservatives and Liberals would have 204 MPs.
In Dion’s P3, the results are less proportional. This is due to his small regions with five MPs, which do not help third parties much, and the preferential vote, which again works against third parties and would have helped the NDP. Voters would have elected 145 Conservatives (a bonus of six, compared with full proportionality), 118 NDP (a bonus of 14), 55 Liberals (short 10), 19 Bloc, and one Green (short 10).
As Dion says, his model is “moderate enough to avoid a proliferation of parties.” In my simulation, the Greens have no one to join Elizabeth May. 

Small regions increase political diversity only modestly
By my simulation, of Dion's 73 multi-member districts, 37 elect MPs of three parties and 35 of only two parties. None elect MPs of four or five parties.

Even in five-MP Outaouais—Mirabel, where the Conservatives just missed a seat, with 6,000 more votes they would have taken a seat -- from the Liberals, so still only three parties win a seat. None of the 51 regions with five MPs elect MPs of four parties.
The small regions hurt third parties, as well as fourth parties.
The 14 four-MP regions are not kind to the third party (usually the Liberal Party in 2011). In my simulation, only three of them elected Liberals. In ten of them, only two parties won seats, with no Liberal outside PEI.  In three of them three parties won seats, but one in Quebec elected no Liberal. In two of them NDP voters elected no MP. In one, all four MPs are Conservatives.
All seven three-MP regions elected MPs of only two parties; five of them elected no Liberal.
Even one-third of the five-MP regions elect no Liberal: ten in the West, six in Quebec, and one in Ontario. Similarly, one-third of the five-MP regions elect MPs of only two parties.
Small regions are good for the two biggest parties (“moderate proportionality”). In Quebec, one of those in 2011 was the Bloc. Province-wide proportionality would let 2011 Bloc voters elect 19 MPs, and Dion’s model also gives them 19 MPs.

Green Party Prospects

The Green party needs to come at least third in order to survive to the final preferential count. That happens in my simulation in Victoria--Nanaimo region (with five MPs in the next election), where Liberals voters’ second choices help elect Elizabeth May.
But although the Greens came ahead of the Liberals in three more regions (Okanagan—Kootenay, BC North—Kamloops, and Central Alberta), in none of those regions did they get enough votes in 2011 for a seat. Still, the Green vote will certainly increase once every vote counts.

In the five-MP district of Calgary North, Green voters cast 15,986 votes in 2011. With another 14,200 new voters, they would have elected an MP such as Heather MacIntosh. That’s within reach. In the five-MP Okanagan—Kootenay region, they cast 22,028 votes, and with another 22,900 new voters they would have elected an MP such as Greig Crockett. Similarly, in the five-MP Vancouver region, they cast 17,597 votes, and with another 21,500 new voters, Adriane Carr would have taken a seat from the Conservatives. And in five-MP Calgary South where they cast 16,138 votes, another 21,200 new voters would have elected an MP such as William Hamilton.

In Ontario, in the Green stronghold of five-MP Guelph—Dufferin—Bruce they cast 20,832 votes. With another 25,200 new voters they would have elected an MP such as Ard Van Leeuwen. In five-MP Simcoe—Muskoka they cast 14,777 votes. With another 24,700 new voters they would have elected an MP such as Erich Jacoby-Hawkins.

And in districts where the Greens are competing for a seat with the NDP and Liberals, they could elect an MP in these districts with fewer votes than stated above, if they took some votes away from those parties.

Back in 2008, the Greens got 6.8 percent of the vote, compared with 3.9 percent in 2011. If their vote went back to its 2008 level, would they have won a seat in Dion’s model?
Their best region in 2008 was five-MP Okanagan—Kootenay where they ran ahead of the Liberals, but got 32,298 votes, less than the quota for election. Once every vote counts, the Greens will do better. If they could have raised that to 43,100 with 10,800 new voters, they would have taken a seat from the NDP.  Maybe Huguette Allen would have been Canada’s first Green MP.
In 2008 in four-MP North Nova—Cape Breton, Elizabeth May would have gotten more votes from the rest of the region (the other three ridings), so if she had gotten 33,200 votes rather than the 17,202 the Greens actually got, she would have pushed the NDP into fourth place and taken a seat from them.

A high threshold does not well serve political diversity

This adds up to the same flaw that sunk the Quebec Liberal model in 2006, which also had five-MNA regions. Quebec’s Select Committee on the Election Act recommended on May 31, 2006: “Since the move to reform is motivated largely by a desire to foster political diversity, it is essential to opt for a formula that gives smaller parties a realistic chance to win seats. The goal of political diversity would not be well served then by a regional compensation formula that set the threshold too high.” Therefore, when defining compensation districts, it recommended “that the principle of increasing the political diversity of the National Assembly be respected.”
Fair Vote Canada prepared, in 2005, an assessment of the model proposed for Quebec. On Quebec, FVC said it “provides a very good foundation on which to build a fair voting system, but the current proposal must be greatly improved. With 14 regions instead of 27, a typical region could have six riding seats and four list seats. More seats per district will allow better proportionality, meaning that many more voters can be represented by their first choice.” (Note that a ten-MP region is still only moderately proportional.)

Can a model solve these problems but still follow Dion’s request for moderate proportionality?
Yes: a moderate mixed-member model based on the Jenkins Commission Report in the UK. It is personal, preferential, and moderately proportional: call it P3-MMP.

Footnote: is it like STV?
Some people think Dion’s model looks like Ireland’s STV system. But the key to STV is the surplus transfer, which Dion has left out for the sake of simplicity.  For example, say the result on the first count is, in quotas:
Conservative 2.6
NDP 1.4
Liberal 1.2
Green 0.8

With STV, the surpluses are transferred first. By the time all surpluses have been transferred, the Greens will surely pick up enough transfers to push them over 1.0 and elect an MP. This leaves the Conservatives with 2 MPs.
With Dion’s model, the Green drops first. Based on the EKOS poll, the second count is:
Conservative 2.73
NDP 1.66
Liberal 1.34
The Conservatives get 3 of the 5 seats, with 43% of the votes.

And unlike STV, you cannot rank individual candidates on P3’s preferential ballot. For example, you cannot rank the candidates from your community ahead of the other parts of the region. Nor can you rank the female candidates, or the young candidates, ahead of the others.

Technical Note:
In this simulation I have assumed that two local stars will attract enough more votes from the other four seats in their region to pull their party's vote above the threshold: Ralph Goodale and Kevin Lamoureux.