EKOS was in the field from February 1-10th with the following headline results with changes from a poll they conducted at the end of November:
BC NDP 39.0% -2.2
BC Liberals 27.4% +3.5
BCCP 14.6% +3.2
Greens 13.5% -5.7
Others 5.5% +0.8
The sample was 687 people of which 576 were decided voters - 83.8% decided. I have no details of the November poll other than the headline results.
I am not sure what to take from this poll, I would like to see at least one more from EKOS soon so that I can see in the detailed tables what their trends look like. The numbers in this poll are significantly different than I would expect.
The EKOS numbers show the NDP lower than anyone else has in the last two years. That seems odd to me. The gap between the NDP and Liberals remains over 10 percentage points which is broadly consistent with other pollsters.
The results for the "other" is higher than the other pollsters have been finding. When I crunched the numbers on what the independents might get in the election, 4% strikes me as the upper limit. The higher result here might be a none of the above reaction. In the 2011 federal election EKOS had among the highest "other" in their polling.
The Green results look a bit high compared with other recent polls. EKOS seems to get better results for the Greens than others do. In the 2011 Ontario election they had the Greens at 5.9% to 7.6% and the party managed 2.9%. In the 2011 Federal election EKOS consistently had the Greens at higher support than most of the pollsters and significantly more than the election day results.
If the high results for the BC Conservatives and Greens are truly accurate they are important because these parties are now at popularity levels were one should expect them to win seats. It is at over 12% support that one should expect a political party to win seats i a BC election. Only the BC Greens in 2001 achieved more than 12% of the vote and no seats but that was a weird election because the BC Liberals took 57% of the vote.
2 comments:
I lie my way through polls and so do a lot of other people.
This is something I have tried to work out the impact of. At the moment all pollsters assume that 100% of the public answering is telling the truth because they have no mechanism to determine who is telling the truth and who is not.
This a source of error in polling that is not accounted for. When polls are reported we should hear about three different margins of error
The Statistical margin of error - this is the one tha tis reported and frankly the least relevant to the results in many ways. Better would be if all the results were published as graphs with bell curves of the results.
Rounding error - rounding to the nearest whole number adds a +-0.5 pp error to each result
Methodological errors - this has to be estimated but it would include people that lie, problems with the survey process, bias in the survey, impact of weighting and more. I think this alone could be as high as +-6 pp i nmost polling in Canada
Post a Comment