News

The headlines contradict, accusations and dire predictions fly, party war rooms are frenetic, and the polling graphs now come in five colours. But an analysis by the Canadian Opinion Research Archive at Queen's University indicates the first two weeks of the election campaign produced plenty of hot air and burnt aviation fuel but little change in voting intentions.

Oh yes, there have been a few upticks and a few downticks, mostly depending on the luck of the samples draw and a bit of a puff for the Conservatives around the election call. But when we take the combined efforts of Nanos Research (12 tracking points since the election writ), Harris/Decima (10 tracking points as of the weekend), Ekos Research Associates (eight since the writ, two just before) and benchmarking from Environics and Angus Reid Strategies, the electorate is just about the same place it was when Stephen Harper crossed over Sussex Drive and asked the Governor General to dissolve Parliament.

That is: The Conservatives are around 37 or 38 per cent of decided voters, the Liberals are somewhere between 25 and 30, the New Democrats still around 16 or 17 per cent and still trying to crack their 20 per cent ceiling, the Bloc Quebecois is between 6 and 8 per cent on a national basis (but since it only competes in Quebec, assigning it a national number is something of a mathematical abstraction) and the Greens still under 10 per cent.

For comparison, the 2006 election results were Conservatives 36 per cent, Liberals 30, NDP 17, Bloc Quebecois 10 and the Green Party four.

There have been some adjustments — the Conservatives have moved up a notch over their 2006 results to a new plateau, with a corresponding drop in Liberal strength.

The Bloc has lost a significant slice of support in Quebec since the 2006 election but there has been little movement during the campaign, and the Greens have increased their salience, maybe even doubled their support, though over a very small base. But almost all of this adjustment took place before the campaign began. So what about all of the headlines, the hours of television footage and commentary in the past two weeks? So far, it would seem they haven't yet counted for much on the ground.

The polls do indicate, however, that Mr. Harper's Conservatives got a lift in late August and early September. The extensive speculation about the impending election call seemed to shake voters out of their summer torpor and triggered an enhanced appreciation of Mr. Harper and his party.

According to CPAC pollster Nik Nanos, who rates party leaders on trustworthiness, competence and vision, the election lead-up permitted Mr. Harper to convey an air of decisiveness and determination at a time when he and his party had the public limelight all to themselves. Note that the increase in Conservative/Harper strength also coincided with a heavy purchase of pre-election advertising. Perhaps the image-modulation ads worked.

Mr. Nanos attributes this initial bump to the "halo effect," and says the numbers tend to move back to more traditional levels in a few days as the reality of the election sets in.

Indeed that seems to be happening if you look at the catch lines the pollsters attach to their releases. On Sept. 14, the Harris/Decima headline was "Conservatives ahead by 14." But the next day, the Harris/Decima headline was "Poll suggests Conservative lead softening." And the same day that Harris/Decima was measuring a gap of 14, Nanos said the gap between the two leading parties was only eight points.

By Sept. 17, Harris/Decima was saying "Majority is still in sight" but by the next day the tag line was "Conservative support inches downward." During this period the Conservatives' support measured by Harris/Decima ranged between 38 and 41 per cent. During the same period, Nanos had the Conservatives stable at 37/38 per cent and Ekos was in the 35 to 38 per cent range.

How can all these polls be right? All of them come with margins of error — estimations of the likely range within which the "real" value falls, based in part on sample size. So two random samples of 1,000 respondents can reveal slightly different figures, and still both be right — that is, both can be within a reasonable range (say, plus or minus three per cent), given margins of error.

In short, taking into account margins of error, 35 per cent and 39 per cent may not be "statistically" different. As a consequence, much of the recent electoral volatility is more in the eyes of the beholder than on the ground. And this is true for polling-based seat projections as well, of course.

EKOS has suggested volatility in the four seat projections it has constructed since the election was called, based on its rolling poll. The tagline on the first, released Sept. 10, was "Tories on the brink of a majority" and the projection indicated 156 seats for the Conservatives (two more than the 154 needed for a majority) and 82 for the Liberals. In the projection five days later, the Conservatives were down to 147 seats. The next day's projection had the Tories back up to a majority with 161 seats. By the start of this week, the model was projecting 151 seats for the Conservatives.

EKOS president Frank Graves made a questionably profound inference based on the minor shifts his company had detected: "The Tories are once again in majority territory, which is much more significant and impressive than the majority position they enjoyed at the outset of the campaign. The electorate have weighed and considered this possibility, and it no longer appears to be having the same deleterious impacts on Tory support that we have seen in the past." Normally, Mr. Graves is more cautious about making grand generalizations on the basis of modest numbers.

Regional breakouts for Ontario and Quebec indicate more churn and more differences between the various surveyors. In the week leading up to the writ, Darrell Bricker, who heads up the political side of Ipsos, declared "the Conservatives have failed to make inroads in Quebec where the Bloc Quebecois remains strong."

Not quite so, according to his competitors. In both Nanos and Harris/Decima surveys, the Bloc is still marginally ahead or tied with the Conservatives in Quebec. But both pollsters indicate that at the beginning of the campaign, the Bloc's vote share had dropped at between five and 10 points since the 2006 election, when it got 42 per cent of the provincial vote - a shift that is enough to translate into seat losses if it is sustained until Oct. 14.

In Ontario, Nanos maintains that the Liberals and Conservatives remain statistically tied. But in a regional breakdown based on the accumulation of five days of nightly tracking, the overall Ontario totals indicating the tie mask substantial differences around the province. It has the Conservatives leading in Southwestern, Northern and Eastern Ontario while Liberals retain their strong lead in Toronto and the surrounding 905 area.

But as sample sizes decrease, as they often do in polls of provinces or swing ridings, margins of error increase. The Harris/Decima survey released on Sept.13 had the Conservatives 10 points ahead of the Liberals in Ontario, while Nanos had them tied. Even a good portion of this gap might be accounted for by rather larger margins of error that come with provincial samples often below 500 respondents.

The bottom line is that two weeks in, despite the minor contradictions and competing claims, election polls have yet to capture a "statistically significant" breakthrough or dramatic change in momentum.

Stuart Soroka is director of the Canadian Opinion Research Archive at Queen's University School of Policy Studies. Hugh Winsor is a fellow at SPS and former Globe and Mail columnist.

Special to The Globe and Mail