In May, 2011, the Conservative Party of Canada stunned political pundits by winning a solid majority government of 166 seats – 13 more than required to control the federal government for four and a half years. Instead of winning a minority government as projected, the Conservatives won 53.9 percent of seats available with only 39.6% of the vote. A five-party vote split and Canada’s first-past-the-post style of electoral politics has allowed for a party with only a plurality of votes to control the majority of politics for the first time since 2000.
Electoral reform in Canada is, sadly, not a seriously-considered issue, but that doesn’t mean it’s not necessary. The Liberal Democrats in the United Kingdom made electoral reform part of their platform, and their preferred choice, instant run-off voting, was put to the test recently in the UK. Though defeated in the UK, I was curious what would have happened in Canada in 2011 if we used the same system. After a little work, I was able to come up with some answers to that question, but before we get to the exact logistics, let’s take a peek at how things work now, now instant run-off voting would work, and the methodology of my thought experiment.
Canada currently uses a Westminister-style parliamentary system, in which the de facto head of state is the leader of the legislative branch of government. Members of Parliament (MPs) are elected from individual districts – or ridings – that represent single, divided sections of Canada. No MPs overlap in their representation and there are no “at-large” districts, except for in the three sparsely-populated territories. We use a “first-past-the-post” (FPP) system for electing our MPs. FPP is very simple: the person with the most votes wins, regardless of the actual percentage of people who choose that particular candidate. This works very well in systems where only two parties exist, as the winner of the FPP race tends to have a majority, but it has led to increasing criticism that it is not often representative of the majority of members of a district.
Consider, for instance, the breakdown for the riding of Richmond-Arthabasca in Quebec:
|New Democratic Party||17,316||32.5%|
|Conservative Party of Canada||13,145||24.7%|
|Liberal Party of Canada||3,711||7.0%|
|Green Party of Canada||1,098||2.1%|
Although under FPP, the district was won by Bloq Quebecois (BQ) member André Bellavance, who won with slightly over one third of the vote. Given that the BQ is an anti-federalist party, criticism of FPP would suggest that he represents the separatist desires of only a third of his electorate, while 66.3% of people in Richmond-Arthabasca voted for a pro-federal party. Regardless, Bellavance remains one of only four members of the BQ left in Parliament, but his views do not represent what two thirds of his electorate selected in the actual election.
|Conservative Party of Canada||21,997||40.4%|
|Liberal Party of Canada||19,128||35.1%|
|New Democratic Party||11,046||20.3%|
|Green Party of Canada||2,159||4.0%|
|Marxist-Leninist Party of Canada||190||0.4%|
Mr. Ignatieff became the first Liberal leader to lose his seat in election since 1945 with this result, but the Conservative candidate, Bernard Trottier, won with only 40.4% of the vote – a clear minority compared to the three progressive parties, who combined for 59.4% of the vote. Yet these electors are now unrepresented. FPP has certainly failed in this riding as well, as opinion polling has consistently suggested that a large percentage of NDP and Liberal electors prefer the other party first.
FPP does sometimes work in Canada, but only in districts where either the populace is overwhelmingly in favour of a particular party (such as the Conservatives in Alberta or the NDP in downtown Toronto). A commonly-mentioned positive feature of FPP is the individual ridings have a single representative, making it easier for an individual person to contact their local MP and get involved in politics. This gives the FPP system a certain folksy charm that matches several Canadian values.
That said, FPP has a special weakness when one ideological wing is divided while the other is united – such as the Conservatives being united in 2011 while the left wing was split between the NDP, the Liberals, and the Greens; or in 1993 when the Liberals dominated the left while the right was split between the Progressive Conservatives and the Reform Party. In this latter election, the Liberals won 60% of the seats with only 41.2% of the vote, while the two right-wing parties split 34.7% of the vote and won only 18.3% of Parliamentary seats. In both cases, FPP has been proven broken, but this broken system is how things are today, and how the Conservatives won a majority government in May.
What’s “instant run-off voting” anyway?
There’s a lot of other ways to vote for someone to run your country, but the one I prefer is instant run-off voting (or IRV). It’s not a perfect system, but I think it matches the benefits of FPP (namely, local representation) while allowing for the majority of electors to select a candidate. Instead of checking a box beside a single candidate’s name, IRV means you rank your candidates in order of preference. If your candidate receives the least amount of votes, then your second choice is considered. If, after dividing all of the first-eliminated candidate’s votes, no remaining candidate has 50%+1 of all valid votes, the candidate with the current least amount of votes is eliminated, and the process repeats. Once someone has achieved 50%+1 of all valid votes, that person is declared the victor.
So, let’s consider my ballot in the last election. I currently live in Central Nova, a safe Conservative riding in Nova Scotia, but I’ll still show you what I saw in May:
If we had IRV, here’s how it would have appeared:
Because Peter MacKay won a majority on the first ballot, my other choices wouldn’t need to be considered at all. However, consider my uncle’s ballot in his riding of New Westminister-Coquitlam, in British Columbia:
We know that the NDP won this riding with 45.9% of the vote, so let’s pretend we’re preferencing some different parties:
When the first votes are counted, we find that the Marxist-Leninist candidate is the first eliminated, and his votes are redistributed. As he only had 95 votes in the actual election, we can presume those votes will not affect the overall outcome heavily. The next candidate eliminated would be the Green candidate; my vote would then be shunted into the Liberal pile for the next count. Presuming nobody had then reached 50%+1, my vote would be counted in the NDP pile the next time out – at this point, the NDP candidate will likely have received enough votes from the Liberals and the Greens to be considered the victor (which my model forecasts would indeed be the case, but more on that later).
So that’s how it works. Some possible drawbacks include the charge that some votes might be counted more than once, and that some votes are not counted at all (vote exhaustion). However, I feel these are minor issues that can be easily explained. Firstly, a vote may not be counted more than once – it may, however, be counted differently as candidates are eliminated. Each person’s preference shifts based on the available candidates. Regarding the possibility of vote exhaustion, this only occurs when a person does not list an order for all candidates, and then their vote would become invalid. I think that ranking all candidates should be a must for a vote to be counted, if I were in charge of changing the system. Regardless, I think this is a significantly better method to choose a Member of Parliament as compared to FPP.
In preparing to figure out a hypothetical map of Canada with IRV instead of FPP, I had to use the final Ekos poll from before the 2011 election to gather “second choice” options. Because IRV is not currently the electoral system we use, Ekos doesn’t gather favourability numbers or anything beyond a second choice. This data is still useful in figuring out how a Canadian would vote if they had a runoff choice, however.
Although the data was not broken down by region, I was able to separate the numbers by recognizing that all the people who support the BQ would live in Quebec. By assuming that equivalent numbers of support would exist across the board (which is obviously incorrect, but was the only assumption I could make with the data given), I was able to extrapolate approximate support within Quebec and without. With those numbers I was then able to begin the process of breaking it down, riding by riding.
The next assumption I had to make was based on the amount of people who chose “no second preference”. At this point, I assume the elector in question would have not filled out the ballot further. Therefore the total amount of votes reduces between counts as I reassign votes from candidate to candidate. If a person’s second choice had already been eliminated, I simply assumed those votes exhausted as well.
Finally, I assumed that those who voted for minor parties had no second choice, and discarded those votes after the first ballot. In situations (like Edmonton-Sherwood Park or Chambly-Borduas) with major independent candidates, I reviewed the situation and made a best guess as how they’d pull support from the major parties based on the ideologies of the individual candidates.
The results of the IRV experiment
So, after I did all the math and set this up, I punched the numbers through a spreadsheet and came up with some totals. Overall, as you’d expect, IRV is bad for the Conservatives. On average, every second supporter of the NDP or the Liberals would throw their vote to the other major progressive party. It doesn’t kick Harper out of government, but he does lose his majority. Both the NDP and Liberals gain seats, while the BQ are reduced to a single seat from four. Elizabeth May retains the seat she won in May, and no other Green candidates come close to being elected.
Here’s the results:
|Party||Vote %||FPP Seats||FPP Seat %||IRV Seats||IRV Seat %|
At a federal level, IRV rewards the Conservatives and the NDP significantly more than it does the other parties. However, the Conservatives are only awarded +7.2% while the NDP are awarded a similar +7.3% above their popular vote; under FPP, the Conservatives receive +14.3% while the NDP had an advantage of seats over their popular vote of +2.8%.
Provincially, IRV seems to fails in certain areas. Consider the vote allocation for Quebec, where the BQ had 23.4% of the vote but would have gained only 1 seat in the province – 1.3% of available Quebec seats. This is attributable to a significant drop in support by Quebecois for the Bloq and a similar strong increase in the NDP, but it seems to lock out a large percentage of the people who chose the BQ. Realistically, the individual breakdown tells of a running battle between the NDP and the Conservatives in Quebec where BQ voters often break in a significant fashion towards the NDP and away from the Conservatives, giving victory to the NDP. The key to remember with IRV is that each riding allows for each vote to count, as we are looking for 50%+1 by breaking down where the vote goes in the end. The final result isn’t supposed to be 100% proportional, but it is supposed to be fair – based on preferences, rather than first-ranked votes.
Some interesting thoughts on the results
The Liberals would be in a good position, compared to their actual results. They’d retain an additional 11 seats, but they would slip further in Quebec to the NDP. Michael Ignatieff would have retained his seat in Toronto. However, two star Liberals would have lost – former astronaut Marc Garneau would have been unseated in his riding of Westmount-Ville-Marie, while Justin Trudeau (the eldest son of former Prime Minister Pierre Elliott Trudeau) would have been defeated in Papineau.
Despite the preferential voting, the Conservatives held on to seats in highly urban areas that I might not have expected – specifically, the ridings surrounding Saskatoon and Regina in Saskatchewan. The reason for that is simple – gerrymandering. Both cities have been split into four districts and then linked into a large surrounding rural area. This has the effect of splitting the left-wing vote four ways, while balancing with a significant rural (and statistically more Conservative) population. All the ridings that were won by the Conservatives that were in the split ridings were won with majorities or very large pluralities, except Palliser, which the IRV model turned to the NDP, and Saskatoon-Rosetown-Biggar, which was a close two-horse race that the IRV model didn’t flip but came within 65 votes of doing so.
Overall, IRV is unlikely to win support anytime soon, likely because it is terribly bad for the Conservative Party. After the long attempts by the Reform Party and their successor in the Canadian Alliance to “unite the right”, a voting system that makes that push antiquated is highly unlikely to be endorsed by their government – and the supporters Harper leaves behind in the Senate whenever he leaves office. It’s only through an alliance of the NDP and Liberals that we’re likely to see IRV even be considered – and alliance that will not occur until the next wave election knocks the Conservatives out of power.
However, I think this thought experiment has been enlightening. To me, it shows the power of what could be if we had a more fair voting system. I think IRV would energize voters, knowing they don’t waste their votes in any system. Getting people out to vote is the key to making democracy work, and a devil’s choice like we see now is a terrible thing. In Canada, if you’re a right-wing person, you have one choice. If you’re a left-winger, you have three. Any surge by one of the left-wing parties tends to be at the expense of another left-wing party – as the NDP took votes from the Liberals in 2011. Some people vote strategically – they vote for the person most likely to win against the candidate they hate. Some people hate wasting a vote and don’t go out at all. Reform would help fix these problems, and I hope these numbers demonstrate some of why.
Note: this is not a proper study. A proper study would have taken significant amount of polling to make feasible. This is an extrapolation of what happened with what might have happened, and isn’t really scientific or ready for publish in a journal. If you have any thoughts on how the methodology might be improved, let me know.