John Diefenbaker, when asked about polls that showed his party headed for defeat, quipped: “Dogs know what to do with poles.” He then went on to defeat. Of course, in those days, polling was easy. They were done live using land-line phones. That was the only option they had and it worked reasonably well. Earlier failures (Dewey defeats Truman) were long forgotten. And so it went. Eventually, we got government by polls.
But in the last decade some people have begun to question the accuracy of polling. The nadir was the failure of the pollster to predict the size of the NDP vote in 2011 and then came the absolute miss of the 2012 Alberta election where a 10% Wildrose lead evaporated as the PCs cruised to victory. Then came the utter mess in BC. The proliferation of cell phones and of Internet based polling, of robo-polls and call display all worked against the pollsters. People increasingly wouldn’t take their calls or participate in their surveys. Cell phones were hard to reach and uncertain as to where the owner lived. It seemed the days of polling was over.
But they kept doing it – and developed techniques to overcome each of those challenges. But can we really trust them?
Yes. And no.
Polls have their limits. Generally reporting of polls has to stop 48 hours before voting starts. Lots of things can happen in 48 hours. Surges that started a few days earlier may peak more dramatically (like the Orange Crush in 2011) or candidates may say foolish things — as happened to the Wildrose. As well, voter turnout may be unexpectedly low or high, throwing off the predictions (in 2006, many people who said they would vote Liberal simply didn’t show to vote leading to the first Harper minority). Young people in particular are uncertain voters so their stated support for one party or another may throw the whole system off.
Still, in eight of the last ten elections in Canada, pollsters accurately predicted which party would win and whether that win would be a majority or minority in 7 of those elections. (In Ontario the Liberals won a few extra seats turning a bare minority into a bare majority). Especially effective were those who aggregated polls — weighting them on the proven reliability of the methods and the size of the sample.
Still everyone is hedging their bets. The polls show that the Liberals will win a majority in PEI today and the NDP will win a majority in Alberta tomorrow. As for the UK later this week, the polls say it will be a hung Parliament with tough bargaining ahead. I think they may have it exactly right in the first and the last. And I’m pretty sure the NDP will win the most seats in the Alberta vote — and a majority is well within reach.
But in the end only one poll really matters, the one on Election Day. If you have a chance to vote this week, make sure you do. Cause if you don’t, you can’t really complain about the result.
And that’s ten minutes.