Polls in Canada generally look at six regions in the country. Four of them are provinces: British Columbia, Alberta, Ontario, and Quebec. The other two are regions, grouping together Saskatchewan and Manitoba to form the Prairies and New Brunswick, Nova Scotia, Prince Edward Island, and Newfoundland and Labrador to form Atlantic Canada. This makes sense, since these two regions have a smaller population than does Alberta - samples for each province are generally too small from which to draw any useful conclusions. But this amalgamation of these six provinces into two regions masks some significant differences.
The current polling averages give the Conservatives 39% in the Prairies against 33% for the Liberals and 22% for the New Democrats. That suggests a competitive two-way race, with the NDP well behind. But the reality is that Manitoba is shaping up to be a very interesting Liberal-Conservative contest, with the NDP trying to hold on to its enclaves in Winnipeg. And with the new riding boundaries, Saskatchewan is a competitive NDP-Conservative race, particularly in Saskatoon and Regina.
Atlantic Canada is looking like a Liberal landslide. The averages give the party 56% support to just 21% for the Conservatives and 19% for the New Democrats. But again that is misleading - the Liberals are doing well throughout the region, but are facing off against the New Democrats in Newfoundland and Labrador and the Conservatives in Nova Scotia.
Let's try and see what the polls can tell us about these regions.
Prairies
First, we'll start with a simple uniform swing between the 2011 election results and the current polling averages to try and figure out how the parties are doing in these six provinces. For the Prairies, we get the following results:
SK: CPC 41%, LPC 29%, NDP 25%, GPC 4%
MB: CPC 38%, LPC 37%, NDP 19%, GPC 5%
This would suggest that the Conservatives hold a comfortable lead in Saskatchewan but are almost tied with the Liberals in Manitoba. And in both provinces, the NDP has fallen to third place.
How does this match up with the polls? In Manitoba, there have been two large sample surveys done in the last few months. A Sept. 19-28 poll by Probe Research surveying 1,002 Manitobans found the Tories ahead with 42% to 32% for the Liberals and 22% for the NDP. A Jan. 16-Feb. 3 poll by Insightrix Research surveying 800 Manitobans found similar results: 44% for the Conservatives, 34% for the Liberals, and 16% for the NDP.
This would suggest that the uniform swing is somewhat understating Conservative support and overstating support for the Liberals, with the NDP about right.
Saskatchewan is more complicated, however, since we do not have any large polls from the province. What we do have are EKOS Research's four polls conducted between Oct. 10 and Jan. 27, giving us a total sample of 193 responses (the individual polls averaged just 48 responses, which carries a margin of error of +/- 14%). If we average those polls out according to sample size, we get 39% for the Conservatives, 31% for the Liberals, and 18% for the NDP. That matches the uniform swing calculations quite closely, at least for the two frontrunners.
But it shouldn't. If the uniform swing is overstating Liberal support in Manitoba and understating Conservative support in the province, it follows that the opposite must be occurring in Saskatchewan. This would suggest that the race is actually closer in Saskatchewan than the EKOS polling and uniform swing estimate would suggest.
That is hard to swallow, considering that the Liberals took just 9% of the vote in Saskatchewan in 2011. Polling done by Nanos Research between Aug. 12-19 and surveying 156 Saskatchewanians gave the Liberals just 25% support in the province, against 30% for the NDP and 41% for the Conservatives (an EKOS poll of 86 people in the province in early July found similar numbers). This is a more intuitive result.
From this, we can conclude that Manitoba is indeed a Liberal and Conservative contest, which the by-election results in November back-up. Saskatchewan is more of a mystery, however. It would be helpful to have some more data for the province - a province that will be an interesting one to watch in 2015.
Atlantic Canada
Let's do the same uniform swing calculations for Atlantic Canada and see what we get:
NB: LPC 50%, CPC 27%, NDP 19%, GPC 4%
NS: LPC 56%, NDP 20%, CPC 19%, GPC 5%
PEI: LPC 68%, CPC 24%, NDP 5%, GPC 3%
NL: LPC 65%, NDP 22%, CPC 11%, GPC 2%
This shows how the Liberals are doing well throughout the region, with the Conservatives most competitive in New Brunswick and Prince Edward Island and the NDP in Nova Scotia and Newfoundland and Labrador.
It is more difficult to assess these calculations because we have seen few polls from the individual provinces in recent months. The Corporate Research Associates released their last set of federal data almost a year ago, from their May 8-30 poll. At the time, CRA pegged regional support at 49% for the Liberals, 24% for the Conservatives, and 24% for the NDP. Not too different from the current levels of support.
In that poll, the only major differences from the uniform swing calculations above were that the NDP was second in New Brunswick and PEI, and the Conservatives were second in Nova Scotia. Support for the NDP was also considerably higher in Newfoundland and Labrador, at 32%.
If we apply a uniform swing from the CRA poll in May 8-30 to the current averages, we get a different picture:
NB: LPC 50%, NDP 24%, CPC 21%
NS: LPC 59%, CPC 26%, NDP 11%
PEI: LPC 67%, CPC 14%, NDP 13%
NL: LPC 56%, NDP 27%, CPC 16%
The truth is probably somewhere in the middle. For the Liberals, it makes little difference - they are at majority support throughout the region. But it does make a difference for the Conservatives and New Democrats. It puts the Tories between 21% and 27% in New Brunswick against 19% to 24% for the NDP. It gives the Conservatives between 19% and 26% in Nova Scotia, with the NDP at between 11% and 20%. In Prince Edward Island, the Conservatives have between 14% and 24% support to between 5% and 13% for the NDP, while the Conservatives sit with between 11% and 16% in Newfoundland and Labrador. The NDP would be between 22% and 27% there. These numbers mean the difference between defeat and re-election for more than a few incumbents.
It would be helpful if CRA reported federal numbers for the region more often, particularly since their overall regional numbers tend to match other polls quite closely. Without that extra data, however, this exercise gives a good idea of why it is better to keep the Prairies and Atlantic Canada as large regions in my own estimations rather than try to guess at the support each party holds in the six individual provinces.