Monday, March 28, 2011

Little relative change in weekend's polls

During these first few days of the campaign, we were treated to a flurry of polls. Most of them were conducted before the fall of the government on Friday, but did include some or all of the post-budget period. So they act as a good starting point for the campaign that has now begun.
These polls by Probe Research, EKOS, Léger Marketing, CROP, and Angus-Reid are all relatively consistent with one another, as well as the polls previously conducted by these firms.

Please note that Probe and EKOS are telephone surveys, while Léger, CROP, and Angus-Reid use online panels. As the campaign goes on and more polls become available, I will track the difference between these methodologies.

For the three national polls, all of the changes from their previous polls conducted earlier in March are within the sampling margins of error, so we can't really say if there has been a national trend. The situation in Ontario is similar.

What we can take from the national results is that it is obvious the Conservatives are in the high-30s, while the Liberals are somewhere around the mid-20s. And aside from the EKOS poll, the New Democrats seem to have jumped quite a bit.

Delving deeper into the regional results, we can see that the Conservatives appear to be on the rise in Quebec. Angus-Reid does have them down four points, but EKOS, CROP, and Léger Marketing all have the party up, and in three of the four polls the Conservatives are running second in the province.

Another consistent trend in this set of polls is that the Conservatives are down in British Columbia in all of them, while they are up in Atlantic Canada. This could be a coincidence, but it will be something to watch for. A Tory drop in British Columbia would put more than a few seats in play.

But generally speaking, most of the national and regional results in this weekend's set of polls were remarkably consistent. Most striking is the result in Saskatchewan and Manitoba, the region that usually has the lowest sample sizes.

Note that there is a report on a Forum Research poll out there, but I won't be including polls where the only information available can be found in incomplete media reports. An email has been sent to the firm, and if I receive the full details of the poll I will include it in my model.

Let's look into these polls individually. The Probe poll is limited to Manitoba, but voting intentions in the Prairies are generally consistent. While the poll has been added to the Prairie projection, it is given a weighting of 54% to represent that it is a Manitoba poll only.

The poll shows very little change from their last poll in December, with the Tories up three points. What is more interesting is the change since the 2008 election. The Conservatives are stable, but the Liberals have captured seven points, most of them (six) coming from the NDP. In Winnipeg alone, the Liberals are up nine points over their 2008 performance. They're running a close second to the Conservatives at 32% to the Tories' 41%.

Within Winnipeg, the Liberals are leading in the southwest, which includes part of Winnipeg South Centre, one of their ridings. The NDP is ahead in the core area, which includes parts of Winnipeg Centre and Winnipeg North. The Tories lead in the rest of the city.

For the EKOS poll, it is worth noting that the Liberals are leading in Toronto with 40.2%, while the Conservatives trail at 35%. The situation is reversed in Ottawa, where the Tories lead with 49% to 33.7% for the Liberals.

The result that stands out in this poll has to be Alberta, where the Liberals stand at 31.9%. EKOS must have thought the same thing, as they polled more people in the province than they normally would, likely trying to correct the result. The projection model awards five seats to the Liberals if the race was actually as close as this, but I imagine it is due to the sampling error. And nationally speaking, the Liberal gain in Alberta from EKOS's last poll is only worth 1.5 points.

EKOS assigned "ceilings" to the parties by combining their first and second choice results. The Conservatives end up with 45%, the Liberals 43%, the NDP 35%, the Greens 22%, and the Bloc 13%, good enough for roughly 50% in Quebec.

Léger Marketing had relatively consistent results in their poll as compared to their last one in early March, but also looked into whether support was "definitive". 89% of Conservative supporters said that it was, while 85% of the Bloc vote is definitive. That number drops to 79% for the Liberals, 74% for the NDP, and 70% for the Greens.

Léger also asked respondents to assign blame for the election. The result was split, with 39% blaming the government and 45% blaming the opposition.

In Quebec, Duceppe is the leader that respondents placed the most confidence in, at 32%. Layton followed with 26%, while Harper (15%) and Ignatieff (6%) were well behind. When it comes to who would make the best Prime Minister, Stephen Harper topped out at 38%, followed by Jack Layton at 21% and Michael Ignatieff at 13%. Angus-Reid asked the same question, and got 32% for Harper, 17% for Layton, and 11% for Ignatieff.

In the CROP poll, the most interesting result comes in the Quebec City region. CROP has the Bloc leading with 36%, up two points from their mid-February poll. The Conservatives are down four to 33%, while the NDP is up four to 22%. It does put a few more Conservative seats in the capital on the bubble.

Angus-Reid's national numbers have been remarkably stable since January. When the polling firm asked its respondents who was best suited to handle various issues, Stephen Harper led on the issues of crime (37%) and the economy (39%).

Jack Layton led on health care (28%) and ethics/accountability (21%), while Elizabeth May led on the environment (33%).

Significantly, Michael Ignatieff didn't lead in any of the categories. He scored best on the economy, though, at 17%.


  1. Thanks for collecting all this info Eric. I don't think either the Liberals or NDP are worrying the Tories in BC as the lead for the Tories is HUGE.

  2. Any luck with the Forum Research poll in getting the full breakdown? I'd love to see how they made their seat predictions.

    I also wonder if the non-reporting of the GPC was because they didn't prompt/offer it as a response to a closed question, or if it was an editorial exclusion.

  3. I'd be very surprised if Ekos polled more Albertans to correct a result... that completely goes against random sampling theory and would skew the national results.

    This is the problem with reading regional numbers from a national poll... they don't matter and you can't interpret anything from them because their confidence intervals are so wide... and you can't just combine polls to improve that confidence because each poll has its own methodology and must be considered seperately. I have no idea how you predict riding-by-riding results, but I would guess the method is quite arbitrary.

  4. Kyle,

    Methodology is explained in several places on the site. Look to the right under the LE DEVOIR image for links.

    As to your other points, I'm not sure what you mean about Alberta. Sampling more people in Alberta to lower the MOE is not unusual, and obviously they would have weighted the results at the national level to ensure the right proportion. It wouldn't skew the national results.

    And an average of polls is almost always better than any one poll, so I'm afraid you're mistaken. Nate Silver has shown that and even I've shown that here:

  5. Léger and Angus-Reid are reporting MoEs for online surveys? Isn't that still considered horribly bad form, and against the code of conduct of just about every market research association out there?

  6. Never mind on Léger, just saw their standard MoE paragraph in English and get what they were doing.

    Point still stands on Angus Reid though.

  7. I feel like very little weight should be taken with the Ekos poll. Having the NDP at 4% in Alberta (10 points behind the Greens), 7% in Atlantic Canada and having the Greens at 20% in British Columbia is crazy. These are results that are way outside anything else we've seen.

    The Conservatives are also quite a bit lower than we've seen while the Liberals are polling a fair bit higher. It seems like more of an anomaly but I guess we'll see when more polls come out

  8. Eric, when I add 39, 41, and 43 and divide by 3, I get 41. why would you say the three polls since the budget have the Conservatives in the high 30's. not so. They have the Conservatives in the low 40's. Just sayin.

  9. The three polls in this post are 35%, 39%, and 39%.

  10. On a somewhat obliquely related point, I was having a discussion today about the predictability of election results based on the pre-election polls, and the general conclusion we reached based on the data we could find was that in some elections it has been a great predictor, while in others the numbers have moved significantly during the campaign - the same conclusion you reached. But there was a lot of discussion on what factors determined which case was more likely in a given election. My gut feeling is that the best predictor of the volatility of the electorate during a campaign is the volatility of the electorate during the 6-12 months preceding the campaign, but I don't have the data to test it. Have you looked at that?


  11. No, I haven't taken such a long view. I think you might be right, however.


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.