The dying art of polling

Photo: Contributed - leadnow.ca

Swing ridings, emerging battleground ridings colour coded to help voters in strategic voting.

If you've turned on your computer, TV, smart phone or radio lately, you've probably heard about the latest poll numbers in the federal election.

The NDP is in the lead, the Liberals are making gains. No, wait, the Conservatives are back on top – the numbers constantly as voting day nears. But how accurate are they?

While polling used to be a fairly accurate system Canadians believed in, its accuracy has wavered.

“The industry is starting to pile up some misses,” says UBC political science professor Richard Johnson, Canada research chair in public opinion, elections and representation.

“The B.C. election was a classic example, Alberta in 2012 was another one. The British election this year was a pretty spectacular miss. There is a big cloud of uncertainty hanging over the industry, I would say.”

He says the entire polling method has changed with the decrease in landlines, increase in cellphones and use of online polling.

Johnson says random digit phone number generated polling still exists and pollsters use it for both land lines and cellphones, but people are not nearly as prone to answer as they once were.

“Compliance, or the response rate, the willingness of people – once contacted – to actually agree to an interview has gone down dramatically,” says Johnson. “It is much harder to construct a representative telephone sample than before and it is more expensive than it used to be.”

But, he says, telephone sampling is still done because it is the only real probability sampling technique left, even though it is highly dependant on a diminishing group of people.

“They take only those willing to complete the interview, which leads to a very biased sample, and you have to figure out how to correct for the biases after the fact.”

Johnson says online polling leads to problems of poor representation.

“The big issue, which has always been there, but is an enormous issue now, and all the firms admit to – is that all the forms of sampling now have more bias in terms of representing the population than they used to."

“Increasingly, polling firms have to resort to weighting schemes after the fact to try to bring the samples in line with population characteristics, and that is not a simple process.”

He says the polling industry struggles each election to figure out where it went wrong but the link is not clear.

“It is very hard to detect, when you do a post-mortem, where the industry went wrong,” says Johnson.

“Either there is some big breach between the electorate at large and that subset of the electorate that is willing to answer polls, or large numbers of people simply change their minds, and it is even possible that they are changing their minds as a function of what they read in published polls, but the fact is – we don't know.”

On that note, Johnson notes the impact polls have is changing.

“Their impact may be diminishing as their credibility comes into question ... This reflects the fact our electoral system creates a strategic issue for ordinary voters,” says Johnson.

Meanwhile, strategic voting is gaining steam.

All over social media and the web, popular sites like LeadNow.ca and strategicvoting.ca, encourage voters to mark their ballots strategically.

For the first time, basically in Canadian history, there is an attempt to try to give definitive signals at the riding level of which candidate is best positioned to defeat the government, says Johnson.

This type of polling hasn't been done in the past, as polling at the riding level is expensive.

But why is this election different?

Basically, the Conservatives are too conservative for some voters.

“The Harper government is a really conservative government,” says Johnson. “Once they got their majority in 2011, they began to enact more aggressively their real agenda – which is a very conservative agenda.”