High satisfaction, low approval: why 2 polling firms measure Gallant differently
Is Premier Brian Gallant on track for re-election or is he in trouble heading into a provincial election year?
The next election in New Brunswick is less than nine months away and polls suggest Premier Brian Gallant is poised to secure re-election. They also suggest that he is one of the least popular premiers in Canada.
What should voters make of these seemingly contradictory polls?
The most recent survey by Corporate Research Associates, conducted throughout the month of November, put the Liberals at 47 per cent support among decided voters — a full 19 percentage points ahead of the Progressive Conservatives.
The poll also suggested that 49 per cent of New Brunswickers were satisfied with the government's performance, compared with 39 per cent who said they were dissatisfied.
But another survey, by the Angus Reid Institute and conducted in early December, found Gallant's approval rating at just 24 per cent. That made him the second-least popular premier in the country, after Ontario's beleaguered Kathleen Wynne.
This is not the first time that CRA and Angus Reid have been at odds.
Since coming to power in 2014, Gallant has averaged an approval rating of just 31 per cent in Angus Reid polling, while satisfaction with his government has averaged 49 per cent in CRA surveys. The Liberals have comfortably led in every CRA poll since the last election, whereas Angus Reid has painted the picture of an unpopular premier.
A New Brunswick Liberal spokesperson, Marc Poirier, defended CRA's methods, calling the company "the gold standard in public opinion polling in the Maritimes."
While CRA uses random phone numbers, Angus Reid relies on an online panel survey "which is not a scientifically representative opinion poll."
The polling industry is changing as more Canadians transition from landline to mobile telephones.
Polling that includes mobile telephones, as CRA polling does, has the advantage of assembling a random sample — everyone with a phone has a chance of taking part in the poll — but plunging response rates have raised some concerns about the representativeness of that sample.
Online method successful elsewhere
Some pollsters have moved online as a result, where respondents are drawn from pre-existing panels. Though there is little history of online polling being used in elections in Atlantic Canada, online pollsters have had some success in other provincial and federal elections.
Still, it can be more challenging to recruit a representative panel in smaller markets — like New Brunswick.
CRA chair and CEO Don Mills says respondents on online panels choose to participate, and because they "self-select," they can't be considered a statistically reliable sample of the electorate.
"All methodologies can have limits and gaps, including telephone polls due to the high refusal rates," says Shachi Kurl, executive director of the Angus Reid Institute.
"Ultimately, you have to look at polling versus results, and we've been able to accurately reflect election outcomes more than 95 per cent of the time. We have a very solid track record."
PCs questioned polls in 2014
This isn't the first time political polls in New Brunswick have been questioned.
Midway through the 2014 New Brunswick campaign, CRA was attacked by then-PC leader and incumbent premier David Alward, who said the company had "been great over the years at playing games" to help the Liberals.
Mills demanded an apology and threatened a lawsuit, though he says now that he decided against going to court because the comments were made in the heat of a campaign.
Alward said at the time that CRA had said in the closing days of the 2010 campaign that the race was close or that the PCs were behind, even though they went on to win 42 of 55 seats.
In fact, a poll CRA released six days before the 2010 vote had the Tories 10 points ahead of the Liberals. The numbers were within the margin of error of the actual popular vote on election day.
It's all in the question
But perhaps these apparently contradictory polls are instead showing how different questions can produce different results.
"You have to be very careful when analyzing approval rating numbers and not conflate them with voting intentions," said Kurl. "We have seen plenty of premiers with low approval ratings who have won majority governments."
Mills pointed out that his company's question, which it has used for more than a quarter-century, "is not just about the premier. It's 'satisfaction with the Liberal government led by Premier Gallant.' It's a more all-encompassing evaluation. It's not just the premier but it's the government overall."
The Angus Reid Institute explicitly asks respondents to give their opinion of the premier himself. That is different than asking about satisfaction with the government — one can be satisfied with the performance of the Liberal government while disapproving of Gallant.
Mills said his company has detected this disconnect in polling in Newfoundland and Labrador, where the Liberal government of Dwight Ball introduced a budget of tax increases and spending cuts in 2016.
"The government has extremely high levels of dissatisfaction," Mills said, "yet when we ask which party they would vote for, it turns out the Liberals lead. So they don't like what's happened, but they look around, and of the choices available, they still would prefer the current government that gave them the harsh budget."
Corporate Research polling on whom New Brunswickers would prefer as premier — focusing on individual leaders — might be more comparable. Trend lines in government satisfaction and decided voting intentions track closely. But CRA's preferred premier polling follows relatively closely with Angus Reid's approval ratings for Gallant.
Since the last election, Gallant has averaged 35 per cent as New Brunswickers' preferred choice for premier with CRA, only marginally better than his average approval rating from Angus Reid.
Mills said even in the preferred premier question, Gallant could fare better than with Angus Reid because CRA asks respondents to assess all party leaders at once. That means respondents are comparing him to alternatives that they may find even less appealing, rather than Gallant alone.
"Voters may not be thrilled with their premier," Kurl said, "but that does not necessarily mean they'll vote for another party."
New Brunswickers tired of change?
So what appears to be a contradiction might instead by an indication that while Gallant's personal popularity could be relatively low, New Brunswickers are still largely satisfied with the government and willing to support it for re-election, rather than opting for one of the opposition parties.
Nova Scotia Premier Stephen McNeil was able to secure re-election in last year's provincial election despite low approval ratings (according to Angus Reid) and preferred premier scores (according to CRA).
Nevertheless, the disparity could suggest that Gallant's support is wide but not deep, a potential vulnerability for PC Leader Blaine Higgs to exploit.
But for New Brunswickers to force a change of government for the fourth consecutive election, the polls might indicate change-for-change's sake won't be enough — they have to prefer the alternative.