How people elect parliaments
On Sunday the Conservative Party of Canada elected its new national Leader, using a highly unusual weighted preferential voting system which gave heartland party supporters relatively low vote weights.
Members of the party – established in 2003 from the merger of two conservative parties – have elected Saskatchewan social conservative Andrew Scheer over his leading rival, libertarian conservative Maxime Bernier from Quebec, by the narrowest of margins – 50.95% to 49.05%.
Bernier was the overwhelming favourite in pre-vote expectations, having led in every published poll. During the pre-2015 Conservative government Scheer, now aged 38, had been the House of Commons’ youngest-ever Speaker, elected to the office at the age of 32.
Leadership candidates Andrew Scheer (L) and Maxime Bernier
The leadership selection was based on the votes of around 140,000 Conservative Party members around the nation, counted centrally and declared at a party convention held on the weekend.
13 candidates campaigned for several months for the post (one further candidate had withdrawn), and the party used a form of preferential voting to choose among them. Most votes were cast online, but some were cast in person at special polling places.
During Saturday’s party convention the less successful candidates were eliminated one by one. Rapt party members looked on over several dramatic hours as vote tally announcements were released at the convention centre.
With three candidates remaining, Bernier led with 40% to Scheer on 38%, and third-placed Erin O’Toole on 21%.
Only when O’Toole was the final candidate to be eliminated did Scheer overtake Bernier, who has led for all 12 earlier counts.
The gripping vote count was not based on a system of equal member votes, however. The system was premised on the voting members in each of the nations 338 electoral divisions – termed ridings in Canada – being collectively allocated 100 points for each riding.
Of the resulting 33,800 electoral points, the winner of the count needed to reach 50%+1, or 16,901 points, which is what Scheer achieved only on the final count.
Maxime Bernier (purple) led Andrew Scheer (green) in more regions of Canada, but Scheer led in much of over-weighted Quebec and most of the Atlantic provinces (image: Wikipedia/DrRandomFactor)
Preferencing was ‘optional’, in that members could express preferences for as few as 1 or as many as 10 of the candidates.
In each riding, ballot counters sorted the ballots cast by local members and allotted the 100 available points in proportion to the votes tallied.
On the first national aggregation of the results, Bernier lead clearly with 28.9% of the points, Scheer had 21.8% and O’Toole 10.6%. Four other candidates had won more than 7% of the points.
12 rounds of elimination and transfer of ballots were then needed to reduce the field to the final two contenders.
As the party’s national counting centre determined in turn the elimination of each minor candidate, the riding counters transferred ballots from eliminated candidates to those remaining, and re-allocated the 100 points among those remaining.
The optional preferencing system meant that the ballots of voters who only listed preferences for the eliminated candidates effectively dropped out of the count, becoming ‘dead votes’ (known as exhausted ballots in Australia). This rule therefore increased the relative influence of the remaining ballots.
But more importantly, because the system was not based on equal vote weights for every party member, the influence of the members around the nation varied quite dramatically.
In ridings with relatively few registered members, the participating voters in this leadership poll had relatively high influence. By contrast in ridings with larger numbers of members, which are likely to be the party’s heartland of supporters, the influence of those eligible to vote was significantly lower.
The party has revealed that a total of 259,010 voters were eligible across Canada. (Around 1,000 more were accepted on the final voting day.)
Across the whole 338 ridings, this gave each potential voter a nominal national average of 0.130 points of influence on the election outcome.
In Canada’s most populated province Ontario, the voters has noticeably lower influence. The 114,508 party members registered in Ontario’s 121 ridings meant that each voter had a nominal 0.106 points of influence – only 81% of the national average.
Voters in Scheer’s home province of Saskatchewan were weighted similarly, effectively weighted at a below-par 0.108 points each – 83% of the national average. Those in British Columbia were worth 93% of the average.
But the smaller numbers of party leadership voters in the Atlantic provinces, where the conservative party usually polls relatively poorly, had much more influence. Those in Nova Scotia were weighted at 180%, those in New Brunswick at 209%, in Prince Edward Island 258% and in Newfoundland 449%.
Voters in Quebec were also highly influential, at 364% of the national average.
The 52 registered voters in the far northern territory of Nunavut were worth a massive 1,474% of the average.
But in prairie province Alberta – home to former Prime Minister Steven Harper and arguably the heartland of the Conservative Party – those registered to vote had just 44% of the average influence.
Each of the 52 Nunavut Conservative Party members was worth 33 Albertans in influencing the outcome of this leadership poll.
These weightings would have been adjusted again by the actual poll, since turnout rates would have differed between the provinces. Overall turnout was around 54% of those registered.
If Alberta registered voters turned out at higher than the national rate, their influence on the leadership poll would have been even lower.
Within every province there would have been registration and also turnout variation between every riding around the already divergent provincial averages given above.
What impact all this effective vote weighting had on the result is hard to say, but with a final result with a margin of only 1%, it is entirely possible that Maxime Bernier might have had more supporters nationally.
The order of elimination of other candidates might also have been different, although it seems certain that Bernier and Scheer would still have been the two final candidates.
During the 20th century the three major Canadian political parties traditionally chose their party leaders at national conventions, but they have all moved to public elections of party members in recent years.
In 2004 the Conservatives chose Stephen Harper as their foundation leader under similar rules to those used again this year, but Harper won a clear 56% of the first preference vote, so the result was not in question.
The Liberal Party adopted the same leader selection system for the first time in 2013, but out of seven candidates that year Justin Trudeau, now the nation’s Prime Minister, won a massive 80% of the first preference vote, eclipsing any significance of the riding vote weighting rules.
The New Democrat Party is will also be selecting a new leader in October this year. They will not use riding vote weighting, so each ordinary party member has an equal vote. Multiple votes will be taken once per week, with additional candidates eliminated every week until one winner has 50% of the votes of members still voting.
The October NDP vote will be the third member’s election of a party leader, with the late Jack Leyton elected in 2003, and outgoing leader Thomas Mulcair chosen in 2012.
Update: CBC is reporting that in terms of real votes of Conservative party members, Scheer beat Bernier 62,593 to 55,544, with around 23,000 votes ‘exhausting’ (in Australian lingo) without giving a preference between the two. That’s 44% to Scheer, 39% to Bernier and 16% exhausting.
Australian elections mainly use compulsory full preferencing, which would have forced a decision out of that last 16% (on penalty of invalidating entirely the ballots of those who refused to do so). Based on Australian experiences with preference, that last 16% might indeed have changed the result.
Bernier would have needed to win around 75% of the exhausted/dead ballots. Would he have been able to get that rate? We’ll never know.
Update 1 June: Eric Grenier at CBC has an illuminating update on this story today. He has gone through the riding-by-riding numbers.
It’s not entirely clear from the story if he has access to the raw vote data in each riding or if he is just analysing the riding points numbers.
In any case, Grenier concludes that Andrew Scheer certainly had the most actual votes of the final two contestants: 53% of them to Maxime Bernier’s 47%, a lead of over 7,000 member votes.
But Grenier also estimates that had just 66 votes been different in the right places in close, low-population ridings, they could have swung the result – on points – to Bernier, while still leaving Scheer easily the preferred candidate of those 7,000 extra voters. Had such an ‘inversion’ result happened, the outcome would have left Bernier with a very questionable win.
The significance of Grenier’s finding is that the points-weighting system the party uses can easily misfire if an election is very close. It didn’t so so at this poll, but the party may want to think through its leader election process to anticipate a future close contest.
By the by, the potential for votes and points to end up mis-matched in voting systems like this is very similar to what happened in the 2016 US presidential election, where Hilary Clinton had three million more actual votes, but Donald Trump won the most ‘points’ in close contests in a few crucial states.