MRIA Vue Magazine - November 2014

Page 11

SP ECIA L FEAT URE

touch a poll from a provider that has not agreed to publish full tables within 24 hours of a survey being released. This puts all companies on an even playing field. It must also be noted that the Brits are not obsessed with the margin of error, as we so often are in North America (the stories in British media outlets rarely mention it). For the British press, it is more important to know that the pollster has agreed to abide by a stringent code of conduct than whether the data was collected face-to-face, by telephone or online. This transparency also ensures that the numbers are real, and not subjected to deceptive modifications. The British Polling Council would frown upon a company that offered two completely different electoral predictions based on a “turnout model” that, supposedly, reflects voters more accurately. This is ridiculous to the extreme. Polls don’t measure two electorates: the one that will show up, and the one the pollster “thinks” will show up. Polls measure one electorate: the actual electorate. Of course, claiming that one party is ahead with one of your models and another one is leading under another makes gloating very easy the morning after an election. Still, it makes a mockery of our profession. Imagine if a hockey analyst said this at the start of the Stanley Cup final: “Looking at the players. I’d say the Rangers will win it, but if the Kings are more motivated, they will win for sure.” This prognostication would generate jeers at the local Tim Hortons. But some people in the polling industry think it is legitimate to offer two different conclusions, and then claim that one of them was correct. Canadian pollsters can and should emulate the guidelines followed in the U.K. Doing so will ensure that the media can distinguish true, unbiased research from statistically insignificant data or partisan promotion. It will also eliminate the reliance on pseudo-insiders who never show their tables, and whose claims cannot be verified. Let’s be clear: their only motivation is to boost the credentials of the parties and politicians who are paying them.

Also, we must abolish the practice of issuing two forecasts. It is a slippery slope that might very well lead to ill-equipped spokespeople uttering the words: “Our turnout model did not take into account the fact that a meteorite would fall in Ottawa, so we had it right all along.” If we don’t take time to analyze what recent experiences mean to our industry, we will fail. Pollsters will sit elections out, and the only electoral information available to the media will be coming out of the campaigns themselves. Insta-polls will take the place of properly conducted research in media discussions. Measuring public opinion, when done properly, is supposed to be a disruptive experience. Not a week goes by when I don’t get an email from someone who says that they do not know a single person who would vote for a specific party, that everybody they come into contact with loves pipelines or that nobody in their home supports the government. We need to continue to properly collect the data that effectively debunks these myths and misconceptions – it is an essential component of a pollster’s service to society. Still, unless we agree to hold ourselves to a higher standard, we will continue to see our bona fide work supplanted by the talking points of pseudo-insiders and lobbyists, or half-baked insta-polls that masquerade as research. It will be the society that we are supposed to be serving that will lose in the end.

Mario Canseco is vice president, public affairs at Insights West. Throughout his career he has been responsible for more than 30 electoral forecasts in Canada, the United States and the United Kingdom, including the municipal elections in Calgary and Edmonton in 2013. He can be reached at mariocanseco@insightswest.com.

vue | NOVEMBER 2014

11


Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.