Tuesday, May 19, 2009

Pollster Profile: Angus-Reid

This is a bit of an addendum to the previous post about Angus-Reid. In the future, when profiling a pollster, I will include this information.

I've tried to analyse whether a pollster tends to poll well for one party over another. Now, this is not necessarily a bias. In all likelihood, there is no bias at all. It is merely a reflection of the type of polling system each pollster uses. What I've done is compare a pollster's result to the average result of the other pollsters in a given month. I've limited the analysis to national, Ontario, and Quebec polls because they are large enough that differences can't solely be chalked up to margins of error.

Nationally, Angus-Reid has tended to poll best for the Conservatives. So far this year, Angus-Reid has polled the Conservatives 1.9 points higher than the average. Next on the list would be the Bloc Quebecois, which is polled 0.9 points higher. Then it is the NDP at 0.7 points higher. The Liberals are polled 1.9 points lower than the average, and the Greens 2.2 points lower.

In Ontario, again, Angus-Reid polls best for the Conservatives at 3.2 points higher than average. All other parties are polled lower than the average, the NDP at 0.4 points lower, the Greens at 2.0 points lower, and the Liberals at 2.5 points lower than the average result of the other pollsters.

In Quebec, only the Bloc Quebecois is polled higher than the average, at 2.7 points higher. The Conservatives are polled 0.6 points lower than the average, the NDP 1.1 points lower, the Liberals 1.3 points lower, and the Greens 1.5 points lower.

As more pollsters are analysed, it will be possible to show which pollsters are best for which parties. Looking at these numbers, Angus-Reid tends to be slightly more favourable to the Conservatives, but it is not a huge margin of difference.

NOTE: I am still waiting on the Ontario polling results for the NDP and Greens from the last Strategic Counsel poll. Thankfully, I was promised to be given the information today from a Globe and Mail reporter.


  1. Eric, in statistical terms sustained differences among the pollsters indicate a bias somewhere. This has nothing to do with the pollsters' intentions; it's just a basic property of estimators. And yes, this statistical bias originates in methods used.

    A more interesting question, and one that I do not see a clear solution to, is as follows. The laurels for the "best estimate" prior to each election have gone to different pollsters in turn. Angus Reid had the best estimate in 2008; in 2004, on the other hand, I believe the laurel went either to Strategic Counsel or SES (Nanos). Anyway, if each pollster favours a particularly biased sample, how can one tell which sample (subset of voters) is most likely to participate in any given election?

    Figure it out, and... :)

  2. I suppose we can't, and have to base our assumptions on past behaviour.

  3. I'm not sure that we can't in principle, although we need data that the pollsters (and Election Canada as well) rarely publish.

    Each election cycle has had its particular issues of interest. I suppose people have a greater incentive to vote if the current issues are theirs. Even income and age data, if available, act as proxies on which a likelihood of voting could be estimated.

    I mention this because the American site fivethirtyeight.com seemed to analyze some of this stuff last fall. Of course, they had many, many more polls and other information drawn together.


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.