Stéphane Dion is right: the Alternative Vote Would Not Help Canada

Stéphane Dion says:

“Our voting system weakens Canada’s cohesion. It artificially amplifies the regional concentration of political party support at the federal level. With 50% of the vote in a given province, a federal party could end up taking almost all the seats. But with 20% of the vote, it may end up not winning any seats at all. This is how Ontario appeared more Liberal than it really was, Alberta more Reform-Conservative, Quebec more Bloc, etc.

Exaggerated regional differences

Dion says “This regional amplification effect benefits parties with regionally concentrated support and, conversely, penalizes parties whose support is spread across the country without dominating anywhere. A party able to reach out to voters across the country is disadvantaged compared to another whose base is only in one region.

“I do not see why we should maintain a voting system that makes our major parties appear less national and our regions more politically opposed than they really are. I no longer want a voting system that gives the impression that certain parties have given up on Quebec, or on the West.”

Preferential voting (AV/IRV) won’t help

The Liberal Party of Canada voted in January 2012 to support preferential voting (also known as the Alternative Vote, or Instant Runoff Vote). However, Dion now says “Preferential voting . . . does nothing to correct the distortion between votes and seats and the under-representation of national parties compared to regional ones. Other changes are needed to find a voting system that best fits the Canadian context.”

So I checked the 2011 election results to see if Dion is right. On the votes cast in May 2011, would preferential voting (the “Alternative Vote” or “Instant Runoff Vote”) do anything to fix the problems he's dismayed by?

To get voters’ second preferences, I used the EKOS poll of “which party would be your second choice” taken April 28-30, 2011.

Justin Trudeau and Marc Garneau would have lost

Liberal voters in Quebec are under-represented. They elected only seven MPs, not the 11 MPs their voters deserved.

AV/IRV would have cut them down to five MPs. Justin Trudeau and Marc Garneau would have lost to the NDP, thanks to Bloc and Green voters’ second choices.

Conservative voters in Quebec are under-represented. They elected only five MPs, not the 12 or 13 their voters deserved.

AV/IRV would have cut them down to four MPs. Jacques Gourde would have lost to the NDP, thanks to second choices of Bloc and Liberal voters. 

That’s because AV hurts third parties, and in Quebec, the Conservatives and Liberals have become the third parties.

Liberal voters in the West are badly under-represented, where they elected only four MPs, not the 11 MPs their voters deserved. According to my spreadsheet, would AV have let them elect more MPs?

Not in BC: second choices would not make any BC seat change hands.

Nor in Alberta: same result.

Not in Saskatchewan, although Liberal and Green second choices would have elected NDP candidate Noah Evanchuk in Palliser.

In Manitoba NDP second choices would have let Anita Neville hold onto Winnipeg South Centre.

In provinces where we already saw three-way races, AV would make a few seats change hands.

Stéphane Dion is right

But as to his very justified concern, exaggerated regional differences, Stéphane Dion is right: AV would do nothing to help. In fact, it would hurt more than it would help.

Dion says "There are also Conservatives in Quebec, traditionally "blue," particularly in the regions, who are entitled to be heard. Despite my Liberal allegiance, I am convinced that the general interest requires that Quebec’s Conservatives be able to make their full contribution to the building of Canada alongside Conservatives from elsewhere in Canada.
I want a federal voting system that fully honours Quebec’s rich political culture, of which we are rightfully proud."

But he is right: AV will not do this. It's not even a step in the right direction; ask Jacques Gourde.

Stéphane Dion’s new proposal for electoral reform:

Stéphane Dion is clear as to the need for Proportional Representation. He has launched an opening gambit in what should become a serious discussion among the opposition parties about voting reform.

NOTE: This blog post is out of date, of historical interest only. My current comments on Dion's model are Stéphane Dion's moderately proportional P3 model: how well would it work?


The following is my out of date post:

Dion says:

“Our voting system weakens Canada’s cohesion. For Canada, the main problem of this distortion effect is that it artificially amplifies the regional concentration of political party support at the federal level. With 50% of the vote in a given province, a federal party could end up taking almost all the seats. But with 20% of the vote, it may end up not winning any seats at all. This is how Ontario appeared more Liberal than it really was, Alberta more Reform-Conservative, Quebec more Bloc, etc. During all the years that the Bloc dominated Quebec’s representation in the House of Commons, they never received a majority of votes from Quebec.

“This regional amplification effect benefits parties with regionally concentrated support and, conversely, penalizes parties whose support is spread across the country without dominating anywhere. A party able to reach out to voters across the country is disadvantaged compared to another whose base is only in one region.

“My main concern with all this is national cohesiveness, even national unity. I do not see why we should maintain a voting system that makes our major parties appear less national and our regions more politically opposed than they really are, favouring regional parties at the expense of national ones concerned with reconciling the regional interests of our vast country. I no longer want a voting system that gives the impression that certain parties have given up on Quebec, or on the West. On the contrary, the whole spectrum of parties, from Greens to Conservatives, must embrace all the regions of Canada. In each region, they must covet and be able to obtain seats proportionate to their actual support. This is the main reason why I recommend replacing our voting system.

“. . . Thus, seats would be truly up for grabs in all ridings, even in the most Conservative ones in Alberta and the most Liberal ones in Toronto and Montreal.
 

"Canada is a diverse country. So in the interest of national cohesion, it is preferable that national parties not be at a disadvantage compared to those with most of their support lying in a single region.

". . there are many other voting systems. In fact, ours is far from being the preferred method in other democracies. It can hardly be found anywhere except for the United States and the United Kingdom (but not in Scotland or Wales).

Seats do not match votes

Dion notes “the major problem with this voting system: the way it distorts the results between votes and seats. This distortion is often significant, creating enormous gaps between the number of seats won by the parties and the number of votes received. There are even times when the party that won the most seats and formed the government did not even receive the most votes. This has occurred in the provinces (three times in mine) and in the 1979 federal election.

“This distortion effect is particularly difficult to accept when a majority government elected by a minority of voters forces the country on an ideological course that is contrary to the preferences of the majority.

“Sometimes, this distortion effect ends up depriving the opposition parties of enough seats to be able to function properly. In one instance in New Brunswick, the opposition did not win a single seat!

“Canada is a diverse country. So in the interest of national cohesion, it is preferable that national parties not be at a disadvantage compared to those with most of their support lying in a single region.

“Of all the democracies, Canada’s Parliament is one where women are not as well represented as they should be. We need a voting system that helps correct this under-representation and promotes adequate minority representation.

“Like so many other democracies, Canada has seen a drop in voter participation in the last few years. A new voting system is required to help us curb this rise in abstentions.

“Preferential voting .. . does nothing to correct the distortion between votes and seats and theunder-representation of national parties compared to regional ones. Other changes are needed to find a voting system that best fits the Canadian context.”

“Preferential voting is a step in the right direction.” But he explained this at the Fair Vote Canada Conference May 26: it is a foot in the door within the Liberal Party to start the discussion on electoral reform. He knows preferential voting has never led to proportional representation anywhere in the world.

Dion is open to other formulas

"I may not have come up with the best formula, and I do keep an open mind. However, it is in this spirit that we need to work to improve our democracy. . . My hope is that the LPC, and all other political parties on Canada’s federal scene, will one day adopt these views, if not the proposition that I submit for discussion."

These points are more important than the details of his unusual model.

Dion’s model: Swedish-style list PR with small regions

Dion advocates Swedish-style list PR, but with small regions like Spain's but even smaller. You elect between three and five MPs from multi-member districts, with some exceptions like northern seats, but generally, as close to five as possible.

He wants open lists like Sweden, and with small regions this would be very feasible. You would have two votes: one for a party, one for a candidate on the party list.

Also, he wants voters to rank the parties with a preferential ballot. Not STV, where a surplus is transferred; just preferential. A party would be dropped from the count in favour of your second choice party if it didn't get enough votes to elect someone in your small district.

He writes “It would produce a fairly meaningful proportional representation that greatly reduces the distortion between votes and seats as well as regional amplification, but at the same time is moderate enough to avoid a proliferation of parties and retain the possibility of a majority government formed by a single party.”

Let’s see if that is true.

1.     The regional amplification effect continues

Liberal voters in the West have been robbed of their voice by winner-take-all for 40 years. Last May the four western provinces elected 72 Conservatives, 15 New Democrats, only four Liberals, and one Green. A normal proportional system, such as the one recommended by the Law Commission of Canada, would have let Western voters, on the votes cast in 2011, elect 51 Conservatives, 26 New Democrats, 11 Liberals, and four Greens.

I did a simulation with Dion’s model on the votes cast last May, with 73 districts as close to five as local geography allows. (My projection is based on the second choice data from an EKOS poll taken April 28-30, 2011.) In the West, Dion’s model would have helped the top two parties, not the Liberals: 57 Conservatives, 30 New Democrats, only five Liberals, and no Greens at all, not even Elizabeth May on the votes actualy cast in 2011 (but she would have attracted more regional votes).

Dion writes “Despite my Liberal allegiance, I am convinced that the general interest requires that Quebec’s Conservatives be able to make their full contribution to the building of Canada alongside Conservatives from elsewhere in Canada.” Yet he proposes a model which, in his small districts, on the votes cast in 2011 would likely have elected only one more Conservative MP, 6 MPs, instead of the 12 their vote share should give them. Conservative voters in Greater Montreal were especially silenced by winner-take-all: zero MPs when they deserved four. Dion’s model would give Greater Montreal Conservatives only one MP.

Liberal voters in Quebec were also robbed last May: they elected only seven MPs, when their vote share deserved 11. Dion’s model would have given Quebec Liberals only one more MP. But Quebec’s second party, the Bloc, would have flourished: 18 MPs.

2.     Why not MMP?

Dion sets up pure proportional representation with no regions (Netherlands and Israel) as a straw man – “I would prefer to keep our voting system rather than adopt pure proportional representation” – while never once mentioning the Mixed Member Proportional option recommended by the LawCommission of Canada, by the Charest government in Quebec in 2005, by the New Democratic Party, by the Mouvement pour une démocratie nouvelle, and by Dion himself in 2006.

Back in 2006 he favoured the Mixed Member Proportional system, with two-thirds of the seats being elected as they are now and a further one-third compensatory (top-up) seats with a five percent threshold in every province, very like the Law Commission.

Dion notes the advantages of having a local MP with “a riding where they were elected, to which they are accountable and on which they depend for their re-election.” But only 45% of Canadians live in cities as big as Regina, Saskatchewan which has 193,000 people.

Yet Dion proposes a model which would deprive 55% of Canadians, those who live in single-MP communities, of a local MP, unlike the Law Commission’s model.

Dion notes that, in the Netherlands and Israel, “voters lose “their” MP and “their” riding. They vote only for party lists.” Again, he never mentions that the Law Commission’s model lets voters vote both for a local MP and for a candidate on a regional list, not just for a list. And he never explains why he has changed his mind.

3.     Small districts.

If we had used province-wide totals with full proportionality the results on the votes cast in 2011 would have been 126 Conservative, 94 New Democrats, 59 Liberals, 18 Bloc, 11 Greens.

In Dion’s model, due to his small districts, I calculate the results as 128 Conservatives, 115 NDP, 47 Liberals, 18 Bloc, 0 Green (but likely 1 Green, and likely 1 more Liberal).

He says “The party that gets the most votes in a riding would probably win three seats out of five or two out of three.” Indeed, based on the votes cast in 2011, I project more than half of the 73 districts – 44 districts – elect members of only two parties. In 20 three-seaters, we find a 2:1 result 19 times out of 20. In 18 four-seaters, only seven are divided between three parties. Even in the 34 five-seaters, 13 of them elect members of only two parties, and the party with the most votes wins at least three out of five seats 59% of the time.

Dion says his model is “moderate enough to avoid a proliferation of parties and retain the possibility of a majority government formed by a single party.” Indeed; it favours the top two parties.

Dion’s preferential ballot would help Liberal voters in Ontario, letting them elect them 11 more MPs, still five less than they deserved. Yet Dion’s model would cost Liberal voters four MPs in other provinces, compared with who they elected last May. Not even one from Saskatchewan; even in a Regina-Estevan five-seater on the votes cast in 2011 the Liberal Party would have been eliminated after the second count with 0.85 quota.(Except that Ralph Goodale would no doubt have attracted more votes from the wider area, still winning a seat.) One less in Nova Scotia, one less in Newfoundland and Labrador, and one less in PEI. Only one Liberal MP from Alberta. Only one more from Quebec, one more from Manitoba, one more from New Brunswick, one from the Yukon, and one from Labrador.

Not even one Green MP. Vancouver Island has six MPs, which have to be two three-seaters. Dion uses the Droop quota. In a Victoria three-seater the quota calculation would have been:

New Democrats 1.35 quotas
Conservatives 1.34 quotas
Greens 0.92 quotas
Liberals 0.39 quotas

After the Liberals’ second choices, we get:

New Democrats 1.56 quotas
Conservatives 1.39 quotas
Greens 0.97 quotas
 

But Dion’s model then eliminates the Greens, making the final count:

New Democrats 1.86 quotas
Conservatives 1.53 quotas

So the three MPs for Victoria would have been two NDP, one Conservative, on the votes cast in 2011. (In reality, May would have attracted more votes from the rest of Victoria, and would have still been elected.)

Charest's 2005 Quebec proposal

This is not the first time a Liberal in Quebec has proposed a "moderate" small-region model. It's what Charest's government proposed in 2005. He proposed an MMP model with five-MNA regions (three local, two regional.) Except in rural areas it would be three-MNAs (two local, one regional.) Pretty similar.

They held public hearings in 2005-6, by a joint committee: a Select Committee of the National Assembly sitting together with an 8-member Citizens' Committee (an excellent consultation model, by the way.)

The superb Report of the Citizens' Committee reflected the public reaction: such a small-region model was not sufficiently proportional to "reflect the diversity of ideas that exist in Quebec society." "In concrete terms, the actual threshold for entry into the National Assembly could be between 13 and 17%. Given that one of the objectives of reform is to ensure effective representation of the electorate in terms of equality of votes, this threshold is far from being acceptable."

4.     Preferential party ballots

Dion’s small districts need a preferential party ballot to help prevent a Conservative false majority. A party would be dropped from the count in favour of your second choice party if it didn't get enough votes to elect someone in your small district. If a party gets less than 16% of the vote in a five-seat district, those votes will transfer to those voters’ second choice. Therefore, he calls it his “P3 model (proportional-preferential-personalized).”

This is not as good as Irish STV, because there are no surplus transfers.  In Ireland, in a five-seat district, if a party wins more than enough votes for two MPs but not enough for three, the first step in the counting process is to transfer the surplus to those voters’ second choice, so that no votes are wasted and every vote counts equally. For example, the Green Party in Ireland used to elect 6 MPs with only 4% of the first preference votes because it got a rich harvest of transfers from being the frequent second choice on ballot surpluses.

In 73 of his small districts with 304 MPs, my projection shows 12 districts where the preferential ballot changes the local outcome.

In five districts where the Conservatives and NDP split the seats, Liberal second preferences push the NDP ahead for an extra seat: five-seater Windsor-Sarnia, four-seater Niagara, five-seater Simcoe—Muskoka, five-seater Edmonton North, and three-seater Nanaimo—North Island. Similarly, in five-seater Quebec City, the outcome would be NDP 2, Conservatives 2, Bloc 1, until Liberal second preferences give the NDP a seat from the Conservatives.  Similarly, in four-seater Montérégie-ouest the NDP and Bloc would have elected two MPs each until Liberal second preferences give the NDP a seat from the Bloc. In three districts the Liberals are just short of a quota for a seat -- four-seater Durham Region, five-seater Southwest New Brunswick, and five-seater Longueuil -- until Green second preferences (and in Longueuil Conservative second preferences) push them over the quota. Similarly, in two districts the NDP gets a seat only after picking up Green second preferences: five-seater Mississauga, and five-seater Halton—Guelph. In special single seats, NDP second preferences give the Liberals MPs in Labrador and Yukon.

Still, in 61 districts the preferential ballot made no difference.  

5.     Strategic voting is alive and well

Dion writes “. . . voters should be allowed not only to rank parties by preference, but also to select a candidate. They would choose the candidate they prefer from among those put forward by the party they select as their top preference. In other words, voters would choose only one candidate in the party of their first choice.”

To vote for a candidate, you must make his or her party your first choice. If your second choice counts, it counts for a candidate you had no voice in ranking. If you don’t expect your first choice to win a seat – like almost all Green Party voters – you will be tempted to vote for another party in order to be allowed to rank a candidate. Who said strategic voting was dead?

6.     Referendum?

He writes “Precedent makes holding a referendum necessary in Canada: changing the voting system would require popular support. To get this support, Canadians must be presented with a voting system that provides them with better influence over the political system.”

The 55% of Canadians who live in single-MP communities will mostly vote against Dion’s model since it would cost them a locally-accountable representative.

Note that Charest’s government was ready to introduce PR in Quebec without a referendum, and the NDP is ready too. The Law Commission recommended "The federal government should prepare draft legislation on a mixed member proportional electoral system as proposed in this Report. After drafting the legislation, a Parliamentary committee should initiate a public consultation process on the proposed new electoral system."

Assumptions
 

Of course, these projections assumes voters voted as they did in 2011. In fact, if voters knew every vote would count, more would have voted -- typically at least 6% more. And some would have voted differently, perhaps 18% of them by one study. No more strategic voting. We would likely have had different candidates -- more women, and more diversity of all kinds. Who knows who might have won real democratic elections?

So Dion's "moderate" model might be good for the Greens after all, if they got more votes. Anyway, it's a vast improvement over winner-take-all. But by having no local MPs, it would be much less appealing in single-MP communities than the Law Commission's MMP model.

Other options

The real Swedish model would work better: they allocate 11 per cent of their MPs to top up the regional results, with similar results to the Mixed Member Proportional model recommended by the Law Commission of Canada. Except the Law Commission’s MMP model would let us elect two-thirds of the MPs from local single-member ridings, and the other one-third from regional open lists, so every MP has faced the voters: the regional vote can be cast by marking your ballot with an X for any regional candidate standing on the regional ballot.

I look forward to comments.