Tuesday, January 27, 2015

We're nine months out from the Canadian election and I don't feel like saying anything definitive about it just yet

As far as elections go, the Canadian federal election this year seems very distant—we’ve still got a good nine months to go before the current scheduled election day on October 19. (Which is odd, considering that every two years when we have national elections, the hot question always seems to be about what that election means for the elections two and four and twenty years into the future.) And no one’s paying attention in the US, and definitely not as much as they were paying attention to the 2014 midterms. While understandable, it’s one that Americans should be paying attention to, more so than the British ones. We’ve kind of got this special relationship with Canada. Yeah, Britain, we know we’ve got one with you too, but let’s be real here—long-distance relationships can sometimes seem a little, I dunno, let’s say distant, and besides, Canada makes much better food.

The current situation
While current PM Stephen Harper’s Conservatives have been in power since the 2006 elections, it was only in the most recent elections in 2011 that they have actually constituted a majority of the House of Commons—unlike in Britain, where Conservatives weren’t able to form a majority on their own to elect their leader as Prime Minister, Canada has no such rule. Therefore, despite his caucus comprising only 40% of the House after the 2006 elections, Harper has remained Prime Minister in the 9 years since.

It could very well change, however. The 2011 elections were unprecedented in the history of the Liberal Party, not because of the size of the Conservative victory (the Tories won less than 40% of the vote and only an 11-seat majority) but because of the size of the Liberal defeat.* In years past the Liberal Party of Canada had been referred to as Canada’s “natural governing party” because of its near domination of Canadian politics throughout the 20th century, during which it held power for almost 69 years. It’s the same party responsible for the single-payer healthcare system we in America love so much (*cough*) and (thanks to the party’s historic reliance on the majority-French Quebec as a bedrock of its electoral support) Canada’s official English-French bilingualism.

Its 2011 defeat, however, was a shellacking worse than either major U.S. political party has suffered since 1936. The former “natural governing party” was reduced to a mere 34 seats out of the 308 in the House. This was not all to the gain of the Conservatives, who didn’t win anything resembling a landslide. Instead, it was largely at the gain of the New Democratic Party, which, like Britain’s Lib Dems, had historically been the third-place finisher in every election. That changed in 2011, when the New Democrats won over a hundred seats and became the Official Opposition (that is, second place) for the first time in its history. 

The slightly-in-the-future situation
2011, however, appears to have been a deviation from the mean that looks likely to revert. Having received only 18% of the vote in the 2011 elections, the Liberals had bounced back by the middle of 2013 to a 13-point lead over the Conservatives, once again relegating the New Democrats to third place. Since then the gap has narrowed considerably, with both Justin Trudeau’s Liberals and PM Harper’s Conservatives pretty much tied in the low 30s.

Like in the UK (and in the U.S., for that matter), however, popular vote isn’t directly proportionate to parliamentary results, partly due to the first-past-the-post voting system (where candidates who win only two votes can hypothetically win elections if all of their opponents win only one vote each) and partly due to the differences in voting bases between the two parties. It’s like in the U.S., where Democrats now have trouble winning majorities in the House without winning double-digit margins in the popular vote because a good chunk of their base—urban voters and minorities—tends to cluster in a few districts that end up voting heavily Democratic. The voting patterns are slightly different in Canada, not least due to the country’s liberal bent compared to the U.S.—the slightly rightist, capital-C Conservative Stephen Harper would probably be considered dead center on the American political spectrum—but the idea remains the same.

The models and the takeaway
As a result, like in the UK, I’ve created separate linear regression models for each party based on their results in the modern era, which I’ve defined as every election since 1982, when Canada gained the power to amend its own constitution independently of the UK. Here’s the results of the regression of Conservative seats won on share of the popular vote (just like in the UK models):


Coefficient
Std. Err.
t
P
Intercept
-27.3877
3.5770
-7.657
< 0.0001
CPCVote
1.9884
0.1058
18.801
< 0.0001
R2
0.9806




And for the New Democrats:


Coefficient
Std. Err.
t
P
Intercept
-8.2454
3.1614
-2.608
0.0350
NDPVote
1.1843
0.1782
6.644
0.0003
R2
0.8631




And the Liberals:


Coefficient
Std. Err.
t
P
Intercept
-38.4387
9.5512
-4.025
0.0050
LPCVote
2.2904
0.2871
7.979
< 0.0001
R2
0.9009




And finally, for the Bloc Québécois (BQ), the regionalist party focused on the interests of Quebec:


Coefficient
Std. Err.
t
P
Intercept
-10.2253
4.8284
-2.118
0.0888
BQVote
2.2822
0.4486
5.087
0.0038
R2
0.8381





The R2 values here, especially for the Conservatives—whose seat shares are almost perfectly predicted by their popular vote shares—are even better than those for the UK models. (The power of the Liberal and NDP models isn’t to be discounted, either.)

What should the takeaway from these models be? First, we can interpret the coefficients the same way as in the UK models. Canadian Tories, unlike their British counterparts, have a higher “starting point” than the Liberals, as indicated by the intercept; that is, for equally low shares of the popular vote, Conservatives are expected to have more seats. For example, in the 338-seat House up for re-election, Conservatives with, say, 20% of the popular vote would be expected to win 42 seats; Liberals would be expected to win just 25 at that level of support. But as popular vote shares get higher, Liberals catch up to and eventually surpass Conservatives: at 45% of the vote, the model predicts Liberals winning 218 seats while Conservatives would be predicted to win only 210. (Obviously, the model breaks down if both parties do end up winning 45% of the vote, highly unlikely as that is.)

Like the UK models, there are certain bounds for usable data; one is obviously lower bounds for popular vote share that would produce negative results when inputted into the model. (A Liberal vote share of 15%, for example, would predict a Liberal win of -4 seats—obviously a meaningless result.) Similarly, some popular vote shares, particularly for the BQ, would produce parliamentary overflow—a BQ popular vote share of 20% would predict the BQ winning 120 seats, even though the BQ will likely field a maximum of 75 candidates in the upcoming election, one for every district in Quebec.

Fortunately for the models, current polling suggests that this election is well within the region of reasonableness for the models. The BQ has consistently polled at around 3-7 points, predicting a win of about 3 seats, which is as many as they won in the last election. (They used to win around 50 seats, but it’s likely that NDP’s big inroads into Quebec mean that there’s little room for the BQ to win that many anymore.) Both the Liberals and Conservatives are polling at just over 30%, which puts them both in the 130-140 seat range, with the Liberals slightly ahead. At the same time, the New Democrats, polling at around 20%, appears poised to lose most of the gains they made in 2011, pulling them back down to about 50 seats.

Differences with the UK model
Despite the relatively sparse polling compared with in the UK (where YouGov conducts a tracking survey for The Sun), I'm actually much more confident in the ability of the Canadian models to predict well the outcomes of the Canadian election than I am in the ability of our British models to predict those elections. This is largely because Canadian politics, unlike British politics, do not appear to be in the midst of a great realignment. In the UK, the three largest parties--Labour, Conservative, and Liberal Democrat--are polling at record low numbers. An insurgent UKIP knocked it out of the park in the European Parliament elections last year and looks poised to win plenty of seats (but we're not sure exactly how many) for the first time in its general election history. At the same time, the resurgence in Scottish nationalism following the independence referendum held last year is probably driving up the numbers for the Scottish National Party, which now threatens to wipe out the Labour Party from Scotland--historically Labour turf--entirely. But Canada still appears to be in the same general party system it's been in for the last 150 years, more or less. It's more likely that a model built on historical data will apply to Canada, where things aren't expected to deviate wildly from historical boundaries, than it will to Britain, where the very voting allegiances of the British electorate are in the air.

The other major difference is that there won’t be a graph showing the popular vote averages like there is for the UK election in the sidebar just yet. The election is still many months away, and polling isn’t nearly as big in Canada as it is in the UK. Therefore the graph would look quite boring now—every now and then some steep valleys and peaks for when a new poll comes in, and then a flatline for the next few weeks as we wait for the next poll to come in.

* I say “in the history of the Liberal Party” and not “in Canadian history” because 2011’s defeat was topped three times by losses suffered by one of the direct predecessors to Stephen Harper’s Conservative Party, the Progressive Conservative Party. Most notably, in 1993, the Progressive Conservatives went from a majority of 156 seats… to two. Not two hundred. Two. Seats. With a loss of 98.7% of their seats and a swing of -27 percentage points from the previous election’s popular vote, I believe this earns the Progressive Conservatives the dubious distinction of having suffered the worst ever electoral defeat for a governing party in the history of representative democracy.

No comments:

Post a Comment