Western Standard
email print

Poll Vaulting

In the run-up to the last federal election, the Globe rushed to publicize flawed polling data and analysis unfavourable to Harper's Tories. Why?

Patrick McGee - June 18, 2007

Five days before the 2006 federal election, the Globe and Mail reported the Conservatives had soared to an 18-point lead over the Liberals. Just two days later, though, the Globe ran a front-page story saying support for a Harper majority had rapidly declined; Conservatives had sunk to 37 from 41 per cent, while Liberal backing had risen to 28 from 25 per cent. Three of the Globe's senior writers offered an explanation: Harper had "strayed from his carefully scripted campaign" by responding "to questions about same-sex marriage" and raising "old Reform party concerns about a biased judiciary, civil service, and Liberal-dominated Senate." The article was persuasive, but as Michael Marzolini points out in his recent study, "Public Opinion and the 2006 Election," there was one problem: neither the "surge" nor the "dip" ever took place. Both were the result of faulty analysis conducted by the Globe's polling "experts," the Strategic Counsel, led by Allan Gregg.

"I really didn't like writing that [criticism] because I have a great deal of respect for Allan Gregg as a pollster," says Marzolini, chairman of Pollara Inc., itself a leading Canadian polling firm. "At the same time, when you make easy mistakes that have such ramifications during an election campaign, it's a bit hard to ignore them." Marzolini says "a lack of experience" led to an awkward series of questions that produced the skewed results, but "there was no untoward intention or incompetence" at play. Other observers disagree.

"the Globe was clearly biased against Harper getting a majority," alleges Stewart Kiff, a former NDP staffer who now runs Solstice Public Affairs. "They published polls which enabled the Liberals to make the 'hidden agenda' argument once again," he continues, adding the polls were tainted by "some pretty shoddy methodological work." Kiff thinks the Globe may well have been trying for what he calls a "slingshot" effect.

Marzolini says the slingshot tactic was identified by chief Liberal pollster David Herle, who expected "a drop in Liberal support would motivate a 'Stop Harper' movement." It's an idea formed watching Harper's support decline in 2004 after he boasted about possibly winning a majority.

Significantly, other polling firms reported no real change in the week of the questionable Strategic Counsel-Globe poll: "We saw nothing different . . . there was hardly any change whatsoever," says John Wright, senior vice-president of Ipsos Reid. Wright says there's "no obligation to publish every poll," and when "an obvious anomaly" appears, it's the editors' responsibility to determine if it should be published.

Marzolini's analysis shows that just such questionable data made front-page headlines three times in the Globe, pre-election. In early December--when "voters were paying little attention and opinion was stagnant," as Marzolini puts it--the Globe's front page read, "Liberals Surging in Ontario." The reality: in one regional sub-sample, polls reported a two-point increase in Liberal support. Not quite a "surge" given the five-point margin of error. A front-page headline the following week read: "Liberals Snatching NDP Votes in Ontario," and was followed by a flawed analysis falsely indicating NDP support had fallen from 17 per cent of decided voters to nine per cent in a single night. An honest mistake, perhaps, but "if the poll had been true," says Marzolini, "it would have been the most devastating blow ever to impact a Canadian political party overnight."

Errors are understandable, but Kiff wonders why the Globe consistently uses polls favouring the Liberals. "The obvious answer," he concludes, "is bias."

Certainly, the wide margin enjoyed by the Conservatives in the weeks before the 2006 election caused angst for progressives concerned at the prospect of "scary Stephen" getting majority control. But did the Globe intentionally use flawed data to skew coverage in an attempt to influence the election?

More articles by Patrick McGee