The dropping of the writ signals the start of the election campaign but it also starts an exam period for pollsters.  The major exam question is known: how close did you get to the election outcome? But there are minor questions, which are as important, and include: Will there be a polling controversy? Will the polls be confusing and contradictory?

Election polling has always been of major importance for the industry.  It also tends to be very good at passing the test. In the last two federal elections, the final vote shares were very close to the final week polling numbers. While there have been situations where pollsters have been wrong, this is often explainable in terms of changes in the minds of voters between the last poll and the election. Thankfully, they have been few and far between. Even in those cases, however, there was considerable hand-wringing in the industry and media.

In the current environment, failing the test would have profound implications because (1) there is considerable scepticism about the accuracy of polling; and, (2) we continue to emphasize our election accuracy as the proof of our quality.

Consider that in response to a series of media articles (link here to the first one) in February of this year, the MRIA was not shy in purchasing a full page ad in the Hill Times to respond and state the case for MR. The response was interesting because while it mentioned the professionalism of the people engaged in the research, it latched on the historical evidence of accuracy.

Myth: Marketing researchers often get it wrong when conducting election polls.
Fact: No, we get it right, and we can prove it. Our members’ election polls were consistently accurate for Canada’s two most recent federal elections, and the voters confirmed this (Exerpt from the ad)

Ultimately, we need to ask ourselves on what foundation should market research now be built. With statistical theory only providing a weak foundation, the industry needs to develop a shared framework for talking about the accuracy, reliability and robustness of polls. And, while it is tempting, we cannot continue to rely largely on our past ability to predict election outcomes as the proof of our continued relevance.

We get it right, we pass the exam, until we don’t.