Skip to main content

The Globe and Mail

If you think election polls are broken, you’re asking the wrong questions

Keith Neuman is Executive Director of the Environics Institute for Survey Research.

Another provincial election has come and gone in Canada, and once again pre-election polling is coming under scrutiny for not getting it right. What used to be a predictable track record of accuracy in measuring voter support leading up to election day has now become anything but predictable or seemingly accurate. As the pollsters correctly tell us, trends have conspired to make the science of voter prediction an increasing challenge.

This stems from stemming from several factors: Growing difficulty in engaging representative samples of the public, diminishing resources available to pay for quality research, declining party loyalty (what McGill University professor Stuart Soroka characterizes as "flexible partisanship"), and declining voter turnout, which makes it difficult to know which survey respondents will act on their stated preference.

Story continues below advertisement

This state of affairs prompts renewed questions about the value of pre-election polling, and whether it has a future in its present form, or at all. The tone of this discussion is often one of betrayal, as if suddenly we can no longer trust our weather forecasters to tell us if it will rain or shine the next day. This debate about pre-election polling will not be resolved anytime soon, but there are two important lessons that can be drawn from it.

The first lesson is to place more realistic expectations around what pre-election polling can and cannot do for us. The aim of such polls is to track the horse-race voter preferences of leading parties and to make a prediction of how voters will cast their ballots on election day. This exercise will never be a true science because predicting behaviours from attitudes is inherently imprecise given the many factors that influence what people do.

The true purpose of published election polling is media attention: For media companies to generate inexpensive headlines and attract audiences, and for pollsters to boost their profile for marketing purposes (and for this they invest their own money since media no longer can afford to pay for polling).

The most important truth about pre-election polling is that it offers very short term value. Each poll is important until the next one is published, and once the vote is in they all become irrelevant except for academics and research methodologists. Pre-election polls have no lasting value to governments, governance or society; like a horse race, once the winner has been announced it's all done and everyone moves on. In the end, the significance of polling is largely in how it feeds the media's need for a narrative to make elections newsworthy.

The second important lesson is to pay more attention to the polling that does not take place during elections and which is not fixated on horse-race winners and losers. This includes the surveys that measure citizens' opinions, perceptions, priorities and expectations about the issues that matter to them and to the broader society (i.e. the economy, health care, taxation, transit, immigration). Unlike horse-race polls, this type of public-opinion research can tell us something meaningful about the state of our society, where it is heading (through trend analysis) and what sorts of policies and initiatives the public might be prepared to support.

Public-affairs research is less encumbered with the problems confronting election polling: Rather than trying to predict voting behavior precisely to determine election outcomes, it focuses on attitudes and the population segments holding different viewpoints on issues.

Election outcomes make a definitive statement about what voters want, but they do not tell us why voters made the choices they did. Pundits, parties and pollsters are now interpreting why a plurality of Ontarians chose Kathleen Wynne's Liberals over the opposition parties, but most of it is speculation and much of it coloured by a partisan view. Were the swing voters attracted to the Liberal leader and/or her platform, or repulsed by the Tory plan for smaller government? On this question the pre-election polls cannot tell us with any certainty, and the vote itself is mute. Exit polling can play an important role in helping to answer this question, but this type of research is rarely done in Canada, due to its cost and media disinterest once the vote is in.

Story continues below advertisement

Public-opinion research focusing on issues does not get the spotlight of election polling, but it is being done and deserves more serious attention. In the United States, the respected Pew Research Center publishes a steady stream of high quality research that paints an invaluable portrait of the American people and how the population is evolving over time. In Canada, there is much less of this research but it is there for anyone choosing to look. A handful of leading Canadian research companies conduct this type of public-affairs research, some of which is proprietary (intended only for the eyes of paying clients) and some published for all to learn from. In 2006, Michael Adams founded the Environics Institute for Survey Research as the first non-profit entity in Canada dedicated to conducting and publishing independent research on public policy and social trends, with a mission to do for Canada what the Pew Center is doing for the United States. This type of research has lasting value for all sectors of society who want to know who we are as a country and a people, without the sometimes distorted lens of media hype and partisan agendas.

In the absence of a more serious investment by media to strengthen the quality and rigor of pre-election polling, it is wise for them to demonstrate greater restraint in what they publish and for the rest of us to consume it with more realistic expectations. In the end, it would be more productive to direct our attention to the polls conducted outside of election periods on the broad questions that really matter.

Report an error Editorial code of conduct
As of December 20, 2017, we have temporarily removed commenting from our articles. We hope to have this resolved by the end of January 2018. Thank you for your patience. If you are looking to give feedback on our new site, please send it along to If you want to write a letter to the editor, please forward to