Turning Votes Into Seats By Darrell Bricker

Toronto, Ontario (As Published In The Globe and Mail): Polls on party support are frequently featured in our national media, especially during the run-up to a federal election campaign. But as all savvy observers of Canadians politics know, winning seats matters much more than winning the popular vote. And, with Canada's first-past-the-post election system, popular vote can be a very confusing indicator of seat counts. There's no better example of this than the 1979 General Election in which Joe Clark's Tories defeated Pierre Trudeau's Liberals to form a minority government even though they garnered 4% fewer votes than the Liberals.

So, what's likely to happen at the riding level in the next federal election? The safest way to get an answer to this question would be to conduct telephone surveys of 300 or so potential voters in every one of Canada's 308 ridings the day before the election. That would be 92,400 interviews in one day. No one, including the major political parties, is going to do this given the resources and dollars involved.

Another way to answer the riding question is to create a seat projection model. It's what the political pros use to understand how the campaign is really going.

How do you create a seat projection model? It's based on logic, some suspended disbelief about margins of error and other statistical concepts, and a couple of simple calculations. All you need are the previous election results broken down by riding, a calculator, and some fresh party support results broken out by region.

Then, you assign each riding to a region (usually BC, Alberta, Manitoba/Saskatchewan, Ontario, Quebec, and the Atlantic). Once you've done this, you take the regional party support results from surveys (available from the websites of the reputable polling firms) and adjust the riding results accordingly. Voila, you've created your own seat projection model!

Now, one serious problem with this type of rough and ready model is that it misses the inter-regional sensitivities that are so important in Canadian politics. A great example of this is Quebec where the Liberal Party has a very inefficient vote. The inefficiency is because the Liberals win a few seats by a large number of votes in the Montreal area. This distorts how the opinion polls look and masks the true state of play in terms of riding outcomes. This is a common phenomenon in many parts of the country.

To get around this problem, we've divided the country into 33 regions based on how Ipsos-Reid does its telephone surveys and some political common sense, and then we assigned ridings to each of the regions. We then rolled together the results of all of the surveys Ipsos-Reid conducted since February 15th, which gives us a lot of interviews to work with.

Darrell Bricker is President, Public Affairs for Ipsos-Reid. This article is based on a presentation he made to the Canadian Public Affairs Association on April 8th in Toronto.

For more information on this news release, please contact:

Darrell Bricker
President & C.O.O.
Public Affairs
(416) 324-2900