Politics Insider delivers premium analysis and access to Canada's policymakers and politicians. Visit the Politics Insider homepage for insight available only to subscribers.
After the disasters in British Columbia and Alberta, the polling industry needed a good result in last night's provincial election in Nova Scotia. For the most part, they got it.
The stakes were high in Tuesday's vote not only for the politicians whose jobs were at risk and the hopeful candidates who dreamed of sitting in the Legislative Assembly, but also for the pollsters who had a reputation to rehabilitate. It came as no surprise, then, that a small province like Nova Scotia had three polling firms conducting daily tracking polls by the end of the campaign. Per capita, the amount of polling in Nova Scotia produced for public consumption was more than four times that of the 2011 federal election.
The extra effort was rewarded. These polls accurately depicted the race in that they pegged Stephen McNeil's Liberals as enjoying a wide, roughly 20-point lead over Darrell Dexter's New Democrats, and that the Progressive Conservatives of Jamie Baillie were nipping at Mr. Dexter's heels for the job of Official Opposition.
A simple average of the final polls from the Corporate Research Associates, Abacus Data, and Forum Research gave the Liberals 47 per cent of the vote, the NDP 27 per cent, the Tories 24 per cent, and the Greens 2 per cent. The actual result (45.5 per cent for the Liberals, 26.9 per cent for the NDP, 26.4 per cent for the PCs, and 0.9 per cent for the Greens, at press time) was hardly different.
The daily tracking poll conducted by Abacus Data for the Sun News Network, which was conducted with live interviewers, was the strongest performer of the night. The final result of their tracking poll, gathered between Oct. 3-6 and surveying some 400 decided and leaning voters, was closer than the last polls from CRA and Forum Research. Abacus Data pegged the Liberals at 46 per cent to 28 per cent for the PCs, 24 per cent for the NDP, and 1 per cent for the Greens, meaning that the firm accumulated a total error of 5.1 points (1.3 per party). All results were within the margin of error of the poll.
Interestingly, Abacus also calculated the voting intentions of likely voters. That polling firms did not do so in British Columbia has been identified as one of the main factors contributing to the miss there. This estimate by Abacus Data was even closer to the mark: 46 per cent for the Liberals, 27 per cent for the Tories, and 26 per cent for the NDP, for a total error of just 2.1 points (0.5 per party) – a remarkable performance.
The self-sponsored polling by Forum Research, conducted via interactive voice response on Oct. 7 and surveying a little over 900 decided and leaning voters, was somewhat less successful. The Liberals were pegged at 48 per cent, the NDP at 26 per cent, the PCs at 23 per cent, and the Greens at 3 per cent, for a total error of 8.9 points (2.2 per party). More problematic was that the miss for the Tories and Greens fell outside of the margin of error of the poll.
CRA's daily tracking poll commissioned by the Halifax Chronicle Herald was conducted throughout the campaign but ended on Oct. 3. The final poll numbers released by CRA, in the field Oct. 1-3 and surveying a little under 300 decided and leaning voters, were the furthest from the mark. Assessing the Liberals to be at 47 per cent to 31 per cent for the NDP, 20 per cent for the Tories, and 2 per cent for the Greens added up to a total error of 13.1 points (3.3 per party). The result for the PCs fell outside of the poll's margin of error.
That CRA exited the field five days before the election was held does potentially explain some of their error, but the polling conducted on the same dates by Abacus Data showed much stronger results for the PCs. In fact, with a total error of 3.1 points, the poll by Abacus Data on Oct. 1-3 was considerably more accurate – and even bettered their own final poll of all decided and leaning voters.
Broadly, the polls tended to overestimate the Liberals. An overestimation of the support of the challenging party was also a problem in Alberta and British Columbia. And both Abacus Data and Forum Research under-estimated the support of the governing party, a problem that also occurred in those two elections. The Progressive Conservatives, previously occupying the third spot in the legislature, were apparently the hardest to pin down: Forum and CRA underestimated them considerably, while Abacus overestimated them a little.
In winning the role of the Official Opposition, it appears that the Tories' incumbents were especially able to hold on to their vote, a phenomenon that also helped Jean Charest's Liberals defy the polls in Quebec's provincial election last year.
However, these are mostly nitpicks. The average error in these polls was considerably smaller than what was recorded in Alberta and British Columbia. No polling firm implied that anything but a Liberal majority government was in the cards, and two of the final three polls put the Progressive Conservatives and New Democrats in a statistical tie. Suffice to say, Nova Scotians were well-served by opinion polls during the election campaign. But it may take a few more successes before the public's confidence in pollsters is restored.
Éric Grenier writes about politics and polls at ThreeHundredEight.com.