Pollsters hate me. Which is strange, because I am such a willing participant – and these days it is tough to find people willing to give up the time to answer a few questions about politicians and the state of the country.
The pollsters were right in their prediction of Tuesday’s Alberta provincial election. Right up until the polls closed I had figured they were wrong, again. This time around it was the New Democratic Party that surged ahead in the polls. And stayed there until election day.
Political polling is supposed to be science, not art, but in Canada it seems to have become a hit and miss thing. In the last Alberta election, in 2012, the pollsters were unanimous that the Progressive Conservative Party’s 41 year dynasty was over, that the new Wildrose Party would win and win big. However, as the cliché so beloved by politicians goes, the only poll that counts is on election day. And on election day the Conservatives won another majority. For every election they predicted correctly (Quebec) it seems the pollsters missed on completely (British Columbia).
I know a little bit about political polling, enough to know that I generally distrust any published poll. Before I even begin to consider believing the published results I want to know the methodology, the sample size, how many undecided, the margin of error, the questions asked, the screening process to ensure that the respondent is qualified. These are all important to properly evaluate the results.
There are so many factors involved that pollsters frequently are at a loss on how to do their business. The advent of the cell phone has changed how they call and who they call. How do you factor in age and ethnicity? How much weight can you give to an internet poll? To an automatically dialed one in which a machine asks the questions of whomever answers? In both cases the pollster has no idea at all who is answering the question, whether the person is even eligible to vote in the election.
It used to be that I never got to respond to polls. I would get screened out in the eligibility questions, either because I worked in the media or as someone who has had paid political employment. The suspicion is I might be a bit biased; for the same reason lawyers are not allowed to serve on juries – they know too much.
Recently though I received a phone call from an automated poll. If a federal election were to be held how would I vote? Who did I think would be best choice for Prime Minister? Just a machine asking me to punch one, two or three in response. I could have been the family cat for all it knew.
I couldn’t resist the temptation. I answered, but didn’t feel under any obligation to tell the truth. I have no idea who the pollster was, the sample size or whether this was a public or private poll. I just knew it wasn’t being conducted properly.
Accurate polling requires trained questioners who pre-screen the respondents. Any other methods are cheaper, but the results are going to be less accurate. In theory you can make up for that with a larger sample size to smooth out responses like mine, but I don’t know that that theory has been seen to work in practice.
This time the pollsters got it right in Alberta. Canada is facing a federal election this fall. I don’t know with 100% certainty who is going to win (though I have my suspicions). I do know that the political polling being done will have to be a lot better than it is now for me to believe it really reflects voter intentions.