Sample size in Kamloops Daily News poll big enough

It’s a fact that the only people who agree with political polls are the candidates who come out ahead. The ones running behind have a standard response: the methodology was wrong, the sample wasn’t big enough, or that old goodie — “that’s not what we’re hearing when we knock on doors.”

The Kamloops Daily News published a poll Tuesday showing the Liberals with an edge in Kamloops-South Thompson, while the New Democrats hold a lead in Kamloops-North Thompson. Those results don’t surprise me. This provincial election is going to have less to do with the candidates than with party politics.

The demographics and boundaries of Kamloops South tilt it toward the Liberals, while Terry Lake’s challenge in Kamloops North has nothing to do with his performance as an MLA and cabinet minister but rather with the fact his riding traditionally reflects strong NDP support. He was able to overcome it in the last election when Gordon Campbell was still a popular leader; this time, not so easy.

But I digress. I’ve learned quite a bit about public-opinion polls over the years, both as a candidate and as a newspaper editor. A poll that is conducted under established scientific methodology is usually a pretty reliable snapshot of what the electorate is thinking on the day it’s taken.

Those who say the poll sample of 300 in each riding is too small to be reliable simply don’t understand polling. In a pool of eligible voters, there’s a threshold sample size that makes the survey reliable if the questions are well-structured and proper random techniques are used. After a certain point, adding to the sample doesn’t obtain a corresponding increase in accuracy.

Oraclepoll Research president Paul Seccaspina pointed this out in his comments to The Daily News. “Once you get past a certain threshold the averages do not change,” he was quoted as saying. He also noted that telephone polling is the best way to conduct a survey.

There is, of course, always the “margin of error” disclaimer that accounts for unforeseen variations.

The poll results don’t mean the election is over and done in our two local ridings. The campaign has four weeks to go, and things can change. New issues and revelations will come up day by day. It’s even possible, at least theoretically, that the Liberals could overcome their B.C.-wide popularity deficit and pull off a win.

But there’s another fact about voter intentions borne out by polling — as a campaign progresses, the number of undecided voters declines gradually until about a week before election day. After that, the undecideds — sometimes as high as 30 per cent — make up their minds who to support in a similar ratio to the decideds before them.

In other words, the chances of turning an election around become progressively tougher as the campaign wears on.