Calling Nostradamus: Projecting a Tory majority?

This story about political scientist Barry Kay put me in mind of my colleague Dan Gardner’s excellent book, Future Babble, about the  inability of experts to make accurate predictions in their field of expertise.

Kay is predicting, based on current polling trends, that a Conservative majority government would be the result of an election  held today.

He could well be right. There are only really four realistic possible outcomes of another federal election — Tory minority, Liberal minority, Tory majority or Liberal majority. So, there’s a 1 in 4 chance Kay is right.

But predictions like these are always leveraged on the premise “if an election were held today.” There will not be an election today, so the prediction is merely entertainment. Things could change and almost certainly will. The Tories will go up in the polls, or they’ll go down.

Projecting seat counts from polling results is a notoriously inaccurate art form to begin with. A typical national poll of 1,500 respondents  means about only five respondents for every riding in the country, so there’s no way to gauge trends at the riding level. Instead, extrapolating a seat count requires constructing models based on previous elections, which by definition are dated and largely useless.

But let’s take a look at a prediction from shortly before the 2006 election. In a story published in the National Post on January 17, 2006 — less than a week before the vote — Kay predicted a major Conservative breakthrough in Ontario:

That estimate is in line with the most recent made by Barry Kay, a Wilfrid Laurier University political science professor who collates public opinion polls to project how many seats a party will win. He currently forecasts the Conservatives will take 59 seats in Ontario, compared with 35 for the Liberals and 12 for the NDP.

Prof. Kay sees Liberal support concentrated in the 416 area code in downtown Toronto and in the band of about 35 seats in the 905 area code between Oakville and Oshawa. Most of the rest of the province, other than pockets of NDP support in Northern Ontario and around Hamilton, would turn blue if current polls hold.

His numbers suggest high-profile Liberal losers could include Belinda Stronach in Newmarket-Aurora, Tony Ianno in Trinity-Spadina and possibly even Michael Ignatieff in Etobicoke-Lakeshore.

In fact, Conservatives took 40 seats in Ontario and the Liberals won 54. So, Kay overestimated by the Tory vote in Ontario by 32 per cent and underestimated the Liberal total by a spectacular 54 per cent.

Sure, things could have changed dramatically in the final week of the campaign, and maybe the polling data was flawed, but in any endeavour, missing your target by more than half is a poor performance.

As for the the high-profile losers,  Ianno did lose but Stronach and Ignatieff were both elected.

For what it’s worth.

Leave a Reply

Your email address will not be published. Required fields are marked *