“Indeed! It illustrates why the ritualistic ‘within three per cent 19 times in 20’ is nonsense, when less than 10 per cent of people respond to polls.” – Political Scientist Barry Kay, Wilfrid Laurier University
Professor Kay was commenting via email on four new federal election polls in the past week. Two polls put the Conservatives ahead of the Liberals – one by six points, the other by eight points. The other two had the Liberals ahead – one by one point, the other by six points.
How could political research companies that claim accuracy to within three (in some cases, 2.5) percentage points produce surveys showing a 14-point disparity – from an eight-point Conservative lead to a six-point Liberal lead – all in the same week? What, if anything should we, as prospective voters, believe?
Let’s start with what we should not believe. We should not believe that any poll is accurate within that ritualistic three points just because a polling company claims it is. The 14-point discrepancy demonstrates that some of those four polls had to be wildly wrong. And they are probably wrong consistently.
As Prof. Kay suggests, the declining response rate is part of the problem. In the old days, when the Gallup organization had a virtual monopoly on political polling, a poll-taker with a questionnaire and clipboard would sit down in our parents’ (or grandparents’) living room and record their answers to a questionnaire. The response rate must have been close to 100 per cent. But the in-person poll has gone the way of the Edsel.
These days, pollsters employ a variety of methods to get answers. These range from live surveys by telephone (with a 90-plus per cent hang-up rate) to IVR (interactive voice response – a robocall with recorded questions answered by punching numbers on the phone keypad) to INT (an internet survey, usually based on a panel of voters selected in some manner by the polling firm) to hybrid (a combination of various techniques).
How the method(s) used affects any given poll is mystery to prospective voters.
In any event, the national numbers – Conservatives up by eight or Liberals by six – matter less than the trends, and the national scorecard matters less than the regional splits.
As the late Liberal Senator Keith Davey, the greatest campaign strategist of his era, wrote in “The Rainmaker,” his 1986 memoir, “Analyze trends, not specific numbers. In other words, look at the polling forest and not the statistical trees.”
Davey’s injunction, based on his success with campaigns for Lester Pearson and Pierre Trudeau, is conventional wisdom among campaign organizers today. The national trend for the past few weeks has been away from the Conservatives and toward the Liberals. Andrew Scheer has gone flat while Justin Trudeau is limping forward.
The latest seat projections suggest a narrowing and desperately close race for the 338 Commons seats. The poll aggregator, 338Canada.com, projects the Conservatives to win 149 seats and the Liberals 147. Prof. Kay, using a model he developed for the Laurier Institute for the Study of Public Opinion and Policy (LISPOP) also has the Conservatives barely ahead, 148-145.
The key is Ontario, which has the most seats (121) and the most swing seats up for grabs. Earlier in the year, the Conservatives enjoyed a significant lead in the province. That lead is gone. The trend is clearly to the Liberals – 338Canada.com projects the Liberals will retain 60 Ontario seats while the Conservatives take 48 and the NDP 13. Barry Kay’s projection is almost identical: 58 Liberals, 49 Conservatives, 14 New Democrats.
Trudeau’s Liberals won 80 seats in Ontario in the last election in 2015. For them to win “only” 58 or 60 seats in Ontario would be a significant setback and a personal slap to the face of the prime minister. But it would be a lot better than it might have been. For which, Trudeau might wish to thank Conservative Premier Doug Ford for all his help – be it unwitting or witless.