(Note – A commenter has pointed out I accidentally plugged in the value for Atlantic Canada’s CPC support into Quebec, which threw all my calculations off. I’ve left the original entry, but I’ve added some corrections in red.)
So I looked more closely at the Ekos poll from April 25 today. I discovered that there was a flaw in my interpretation of their data. Ekos reported they had sampled 2783 people; which is fine, and then went on to quote urban areas in a separate part of their tables, along with the sample size coming from that urban area.
Initially, I thought that the chart meant the data was from the province excluding the urban area, and then the urban area on its own. But I was wrong. The data combined, in the percentage reported, the urban area as well.
For example, for Quebec, Ekos reported:
CPC: 28.1% GRN: 6.4% LPC: 13.1% NDP: 38.7%
Correction: CPC should read 14.7%)
Sample size: 716
and for Montreal, they reported:
BQ: 18.8% CPC: 12.6% GRN: 8.1% LPC: 16.3% NDP: 42.6%
So, for the sake of a seat prediction, since I’m mapping the changes to a geographic area, it is wrong to map the entire province (including Montreal) to the ridings outside of Montreal. I have to split out the outlying constituencies from the provincial results, which adjusts the various provincial figures. The outcome is startling.
Remember, nationally, the numbers don’t change:
BQ: 6.2% CPC: 33.7% GRN: 7.2% LPC 23.7% NDP: 28.0%
However, when you split out the cities from the non-urban areas, the data changes drastically:
Quebec (Including Montreal)
CPC: 28.1% BQ: 25.2% LPC: 13.1% GRN: 6.4%
Correction: CPC: 14.7%
NDP: 42.6% BQ: 18.8% LPC: 16.3% CPC: 12.6% GRN: 8.1%
Quebec (Excluding Montreal)
CPC: 36.3% BQ: 28.6% LPC: 11.4% GRN: 5.5%
Correction: CPC: 15.8%
As I expected (and which is what lead me to revisit this poll, and I’ll be using this new information for Ekos polls in the future) the NDP support is mainly concentrated in Montreal, and has increased more modestly outside of Montreal, to the point where the NDP is in a dead heat with the Conservatives. This could lead to vote-splitting between the NDP and the Bloc Quebecois, electing more Conservative MP’s.
I also split out the results for other provinces where Ekos included urban breakdowns:
Ontario (Excluding Ottawa and Toronto)
CPC: 33.4% NDP: 30.7% LPC: 19.6% GRN: 14.0%
Alberta (Excluding Calgary)
CPC: 58.7% NDP: 20.8% LPC: 14.0% GRN: 5.7%
British Columbia (Excluding Vancouver)
CPC: 34.3% NDP: 30.9% LPC: 20.7% GRN: 13.3%
In our First-Past-The-Post parliamentary democracy, it isn’t regional percentage of support, it is the vote count in each of the 308 individual constituencies, and when forming a government, it is the number of seats in the caucus that matters.
Putting in this new information into my seat prediction algorithm, I come up with the following:
CPC: 160 NDP: 94 LPC: 43 BQ: 11
Correction: CPC: 144 NDP: 107 LPC: 43 BQ: 14
Taking the regional Margin of Error into account, we get a range of seats for each party:
CPC: 159-201 NDP: 13-95 LPC: 43-65 BQ: 11-27 GRN: 0-2 (!!)
Correction: CPC 143-176 NDP: 18-108 LPC: 43-67 BQ: 14-45 GRN: 0-2
This puts the CPC
well over the top (oops – it doesn’t) with 159 143 seats minimum, possibly more. It all depends on the vote distribution; but things are looking very bleak for both Michael Ignatieff and Gilles Duceppe. I can’t say I’m unhappy about that…