The election is getting interesting. The NDP have surged in support and are now tied with the Liberals nationally. Furthermore the NDP have moved ahead of the Bloc and are in first in Quebec. In the rest of Canada, the NDP has pulled into second place in every province except Ontario. It will be interesting to see if all those Liberals who for years have told NDP supporters to "strategically" vote Liberal in order to block the Conservatives will now take their own advice and vote NDP en masse. Over half of Liberals say that the NDP is their second choice, and just over 40% of Bloc supporters have the NDP as a second choice. If NDP momentum continues to grow, this could mean a monumental shift in the Canadian political landscape.
With all this talk of polling, I want to comment on methodology and reliability. The major polling firms are EKOS, NANOS, and Angus-Reid. Nanos polls on behalf of CTV/Globe and Mail and runs 3-day rolling polls whose results are published daily. Ekos does a new poll every couple of days, and Angus-Reid seems to be weekly or erratic.
Nanos is by far the most commented on poll due its association with CTV and the Globe and Mail, however it is also the most unreliable. Their methodology is highly suspect in that their random sample is landline based. This makes it extremely hard to get a truly representative sample, and since no one under 40 has a landline these days, is biased toward older people. As a result, Nanos consistently has the Conservatives higher and the NDP and Green Party lower, than in the other polls. This is particularly prevalent with the Green Party, who usually come in under 4% in Nanos polls, while polls conducted in the same time period by the other firms will have them at 7-10%. This discrepancy is well outside of the margin of error, and can be chalked up to the fact that people the bulk of Green Party support comes from people under 25, who are also extremely unlikely to have a landline.
Ekos has a much better methodology, as they are able to call not just landlines, but cell phones as well. For this reason I use Ekos polls for my speculations, in addition to the fact they collect 2nd choice information, and have a larger sample, and therefore less margin of error. Angus-Reid's internet polls have very large samples, and in the 2008 election proved to be the most accurate, however they are too infrequent, and might actually discriminate against older people who are less likely to have internet.
Now since the election is actually turning out to be interesting, in addition to making the usual complaint about the electoral system, I want to try to predict how these polling numbers will translate into actual seats. Usually I just do this by extrapolating poll results and working them into how this would translate to seats in a modern electoral system, and I will do that as well, but I also want to look how things might actually turn out on May 2nd.
So basically I will take Ekos' own seat projections which are quantitative in nature, and look through individual ridings and try to compensate for qualitative factors, like star candidates, local scandals, and things of this nature.
So according to Ekos' quantitative prediction method based on their polling results, they come up with the following make up of the house of commons based on their April 21st poll:
Conservative 134
Liberal 82
NDP 60
Green 0
Bloc 32
Correcting for qualitative factors, I come up with:
Conservative 135
Liberal 77
NDP 58
Green 0
Bloc 36
Independents 2
If the NDP surge continues in Quebec, they could steal 10 or so more seats from the Bloc. They are also within striking distance of a handful of Conservative seats in Saskatchewan and BC.
Now if you factor in second choices, and if Canada had the proposed British system of alternative vote, then, due to the fact that the NDP is by far the biggest 2nd choice you get:
Conservative 114
Liberal 79
NDP 108
Bloc 7
Green 0
Independents 0
If we had a mixed member proportional system with candidate ranking, we would get:
Conservative 142
Liberal 103
NDP 108
Green 33
Bloc 27
Saturday, April 23, 2011
Subscribe to:
Post Comments (Atom)
0 comments:
Post a Comment