Monday, September 14, 2009

New IR Poll: 9% Conservative Lead

Ipsos-Reid released a new poll today, via Canwest. Hopefully the rest of the details will be released tomorrow on their website, as is usually the case.

The poll was taken between September 10 and September 13 and involved 1,001 Canadians. You may remember the last Ipsos-Reid poll, and why it was a little suspect. This one is pretty much a re-hash of the last one, and when you take into consideration my assessment of Ipsos-Reid this morning, you might want to take this result with a grain of salt, for two reasons:

Conservatives - 39%
Liberals - 30%
New Democrats - 12%
Bloc Quebecois - 9%
Greens - 8%

And those two reasons are that Ipsos-Reid polls best for the Conservatives and worst for the NDP. To have the Tories so high and the NDP so low makes this an outlier, unless we have some polls from other firms in the next few days to back this one up.

Anyway, let's go through the information that is currently available.

The race is nevertheless close in British Columbia, with the Tories at 39% and the Liberals at 30%. The race isn't so close in Alberta and the Prairies, with the Conservatives polling an incredible 72%, with the Liberals at 15%, in Alberta and the split 50-23 in the Prairies.

In Ontario, the Conservative lead has actually narrowed since the last Ipsos-Reid poll, but they still lead 46% to 36%. In Quebec, the Bloc is steady at 36% while the Liberals are up at 28%. In Atlantic Canada, the Liberals lead the Tories 39% to 31%.

I'll be able to project the seats for this individual poll once we have the rest of the regional details. I'm not even sure this would get Stephen Harper a majority, what with the results in British Columbia.


  1. And yet, it is within the margin of error with the latest Nanos poll which showed: (+-3%)

    CPC: 38%
    LPC: 33%
    NDP: 15%

    I understand the reasoning behind your 'assessments' of the differing pollsters but fundamentally your assessment is based on the belief that the pollsters are systematically skewed one way or another.

    To illustrate why I feel this is inappropriate I'll give an example. You'll recall that in the 2006 Nanos polls became the 'gold standard' by coming within 0.1% of the actual values. Then in 2008 they were so far off with their final values that they ranked last out of all available pollsters. (Their laughable attempt at justification using a single day of data aside)

    Did Nanos and their methods become less reliable between 2006 and 2008? I don't think so, what happened in 2006 was Nanos became extraordinarily lucky (their margin of error is +-3% so they can't even be considered 'accurate' to the tenths of a percentage point). And in 2008, they were especially unlucky.

    Another example is the attempts by Polling Observatory to account for what they saw as systematic 'bias' in the polling results, which ended up being significantly off.

    I think it better to simply assemble the data without accounting for biases one way or another, but its up to you.

  2. The model doesn't take into account any perceived bias, I just thought I'd point it out considering that Ipsos-Reid has put out two consecutive doubtful polls, by coincidence on the same morning that I point out that they have been the most and least favrouable pollsters for the NDP and Conservatives. This poll represents the highest and lowest Conservative and NDP result in the projection.

  3. Since your focus is polls, you really should learn the definition of "outlier". This poll is not an outlier.


COMMENT MODERATION POLICY - Please be respectful when commenting. If choosing to remain anonymous, please sign your comment with some sort of pseudonym to avoid confusion. Please do not use any derogatory terms for fellow commenters, parties, or politicians. Inflammatory and overly partisan comments will not be posted. PLEASE KEEP DISCUSSION ON TOPIC.

Note: Only a member of this blog may post a comment.