Tuesday, September 2, 2014

Liberals with wide lead in first New Brunswick projection

In ThreeHundredEight.com's first projection for the provincial election scheduled on September 22, 2014 in New Brunswick, the Liberals under Brian Gallant have a commanding lead and are in a position to win a comfortable majority government.

The Liberals currently lead with a projected 47.4% of the vote, or between 45% and 52%. The Progressive Conservatives are well behind with 29.7%, or between 29% and 33%, while the New Democrats are in third with 16.4%, or between 12% and 20%. The Greens are projected to have 5% support.

With these support levels, the Liberals would likely win between 32 and 41 seats and form a majority government (25 seats are required). The PCs would win between eight and 15 and form the Official Opposition, while the NDP would take between zero and four seats.

Note that even at the extreme low, the Liberals still take 24 seats. Virtually every scenario gives them victory, and with a majority government.

Before getting into the details of the projection model for this election, let's take a brief look at the two polls that have gotten it off the ground.

The first poll, from Forum Research, was unexpected. It was published to the company's website, and is the first foray into New Brunswick that Forum has conducted.

The poll found the Liberals to be leading with 46%, followed by the PCs at 31%, the NDP at 15%, and the Greens at 7%. In that it differed very little from the last public poll at the time, conducted by the Corporate Research Associates in May, it was rather unremarkable.

It did contain some approval ratings, however, which have been lacking for New Brunswick's political scene. PC leader David Alward's were the lowest, at 27% approval to 60% disapproval. Gallant and NDP leader Dominic Cardy were tied at 34% approval, though Gallant had a higher disapproval rating (33% to 29%). Both leaders are still relatively unknown, with 33% not having an opinion of Gallant and 37% saying the same of Cardy. For a third-party leader without a seat in the legislature, that is not bad for Cardy. But Gallant's numbers are a little high for the apparently-incoming premier.

CRA found very similar numbers to Forum in the poll they published today. The Liberals had 48%, with the PCs at 29% and the NDP at 17%. For the Liberals, that was a drop of five points since May but the party has been wobbling back and forth for a year. The PCs have been steady as well.

Of the entire sample, 14% were undecided and, in total, 25% were either undecided or did not give an answer to the voting intentions question. While that is a big drop from the last poll, this is due to CRA not including leaners in their polls outside of an election campaign.

Gallant topped the list on who would make the best premier with 35%, followed by Alward at 22% and Cardy at 11%. Satisfaction with Alward's government was just 35%. He can't win an election with that number.

The projection model

The model for New Brunswick uses the same template that has been employed in the last elections in Nova Scotia, Ontario, and Quebec. The basic methodology has not changed at all. But there are a few other differences.

For one, the model is not designed to include regional data. Both the Forum and CRA polls had no regional breakdowns, and it is unlikely that there will be quality regional data available throughout the campaign. For that reason, the model is just looking at the province-wide numbers. If support increases and decreases uniformly, this will pose no problems. If support jumps for one party in one region but not another, there will likely be more mistakes. This is unavoidable, particularly in a small province like New Brunswick. I will endeavour to include any regional data that becomes available.

Another quirk worth mentioning is the riding of Tracadie-Sheila, the only riding currently projected to be won by the NDP. In truth, it probably will not be won by the NDP. The riding is a strong one in the projection because it was previously the riding of Roger Duguay, the party's leader in 2010. He took 34% of the vote in that election on the riding's current boundaries, which was by far the party's best performance in New Brunswick. But Duguay is not running there again, and is no longer leader. Worse, prior to 2010 the riding had no particular history of voting NDP, and will probably return to that pattern in 2014.

So why does the model still show the riding as leaning NDP? The simple answer is that the model is based on past performances, and anomalies like this one can skew things. There is a mechanism in the model, which I am including for the first time in this election, which takes into account a situation like this one, where a leader is no longer running in a particular riding. But even with that mechanism, based on how the vote dropped in other ridings where leaders did not run again, Tracadie-Sheila is still projected to go NDP. But instead of a 20-point edge, the NDP is just nine points up. It will probably be a miss by the model. I could stick my thumbs in to get a more plausible result, but that would turn the model into an educated guess.

Lastly, my thanks go to Kyle H. of the Blunt Objects blog for providing me with the transposed results of the 2010 election onto the new boundaries being used in this provincial campaign.

Thursday, August 28, 2014

Is Ford back in the race?

Yesterday, rumours were swirling on Twitter that the next poll from Forum Research would show Rob Ford in the lead in the Toronto mayoral race. The rumours were half-right, at least according to the tea leaves of the poll published by the Toronto Star this morning.

Apparently, in Forum's two nights of polls on August 25 and 26, Ford was in a "statistical tie" with John Tory on the first night. Combine that with the rumours, and that probably means Ford was ahead on that first night. 

Note: the averages here remove undecideds from the equation, so that is why the results and the recent-highs/lows are somewhat different from what Forum has reported. Forum has not removed undecideds from its polls.

I took a look at this poll in detail for The Huffington Post Canada. Because of the waves the poll is likely to make today, I thought it worthwhile to highlight the piece here rather than post about a different topic entirely. I found a few issues with the results of the poll that might lead one to think we're looking at a blip that will be reset with the next poll. Regression to the mean and all that.

Now, one might accuse me of setting out to find reasons to doubt the poll. Those accusations would be completely accurate.

Whenever a new poll emerges that shows something unexpected or at odds with what other surveys have shown, those of us who look at these things have a responsibility to investigate whether the poll itself is the reason rather than to look for explanations in the real world first. Are these results terribly unusual? Not exactly - Ford is up four points among all voters, five points among decided voters. Significant but not unworldly. While Tory has led in the last few polls now, he hasn't run away with it, and Olivia Chow appears to be dropping. This is also the first poll since Karen Stintz withdrew from the race. It is perfectly rational that a candidate might make a move in these circumstances.

But Ford is not the candidate of rational politics. I can say with confidence that Canada has never seen a political figure like him before - at least one that actually stands a chance to be re-elected to high office. We all know what he has been alleged to have done, denied having done, was confirmed that he had done, has apologized for what he has done, and what he has done again to restart the whole cycle. That he could win is counter-intuitive, so we must investigate reasons for a surge in support in the poll itself before assuming enough Torontonians are willing to give him another chance.

The next polls will determine whether Ford has indeed put himself back into a position to potentially be re-elected. For now, we should proceed with caution.

Tuesday, August 26, 2014

Two new polls give Liberals wide lead

Both Ipsos Reid and Abacus Data reported new national polling numbers yesterday, and both showed the Liberals leading the Conservatives by six or seven points. The results corroborate some of the other recent numbers we have seen. These suggest that the Liberals have rebounded, at the expense of the Conservatives, from what looked like a slump-in-the-making at the end of the spring.

The Ipsos Reid poll, conducted for Global News, was the first we have heard from the company at the federal level since April 17-22. Since then, the Liberals were up five points to 38% support, a jump that is outside the margin of error of a probabilistic sample of similar size.

The Conservatives were down two points to 31%, while the New Democrats were unchanged at 24%.

The Greens and Bloc Québécois had 3% apiece.

Abacus Data's last poll was conducted June 25-July 3. They have not recorded any major shifts since then, but the Liberals were up one point to 35%, followed by the Conservatives at 29% (down two points) and the NDP at 22% (up one point).

The Greens stood at 7% support, while the Bloc was at 5%.

Both pollsters recorded similar gender breakdowns. The Liberals were up by two points among men in the Abacus poll. The margin was four points according to Ipsos. Among women, both firms put the gap at 10 points between the Liberals and Conservatives.

I've highlighted above the three largest regions, but only in Quebec do the numbers look very close. There, both polls have the Liberals ahead of the NDP (echoing the results of the latest big-sample CROP poll) with the Bloc Québécois well behind.

The dissimilarities in British Columbia and Ontario are, on the face of it, important. In B.C., it is either a close Conservative-Liberal contest, or a Liberal lead with the NDP in second. In Ontario, it is either a wide Liberal lead or a virtual tie.

But if we take into account the margins of error (practically speaking, if not theoretically speaking since these are not probabilistic samples) we see that the disparities are really not so significant.

The chart below shows, simply calculated, where the two polls overlap for each of the parties. It gives a good indication of where things stand in each region.


Nationally, the Liberals are somewhere between 35% and 38%, while the Conservatives are clearly in second with between 28% and 32%. The NDP is in third with between 21% and 25%.

The Liberals are probably leading in British Columbia, where they are mostly ahead when taking into account the overlap. The NDP is probably in third, but overall the race is close.

The Liberals also probably lead in Ontario with between 35% to 41% against the Conservatives' 30% to 37%, and are most likely ahead in Quebec, with 31% to 41% against the 22% to 32% of the NDP. In Atlantic Canada, the Liberals are indisputably in front.

The Conservatives are comfortably in the lead in Alberta, but are well off their historical pace. They are also probably in the lead in the Prairies, but it is interesting that at the low end of the Tories' range and the high end of the range for the Liberals and NDP, the three parties are tied.
The Ipsos and Abacus polls would result in very different national scenarios in terms of seats. The Ipsos poll would give the Liberals 150 seats to 121 for the Conservatives and 67 for the NDP - enough to give the Liberals a comfortable minority government.

The Abacus poll is far closer, however, with 135 seats for the Liberals, 122 for the Conservatives, and 70 for the NDP.

If we combine the two results, taking the best and worst numbers in each of the regions, the Conservatives could conceivably come out ahead in the seat count with 140 to 133 for the Liberals. That would be on the margins of what is likely, just as a far larger Liberal victory of 152 seats to 103 for the Tories would be. But with these kinds of numbers, there is a great deal of scope for variation in the seat count. The Liberals need to be over the 40% mark to be in a good position to win a majority government.

The numbers above do give an indication of where the battlegrounds will be. For the Conservatives, much depends on their results in British Columbia and Ontario. For the NDP, it is Quebec and B.C. that are most important, while Ontario will probably make or break the election for the Liberals (I'd throw Quebec into that as well, but the Ipsos and Abacus polls were consistent there for the party).

Also of note is Alberta. With the Conservatives dropping significantly in the province to around 50%, it opens up the potential for a number of seats in Edmonton and Calgary to go to either the Liberals or NDP. It likely won't determine the fate of the government, but the eight-seat range for the Tories above could turn out to be very important. Admittedly, however, if the Conservatives are reduced to 21 seats in Alberta they are likely not doing very well in the rest of the country.

The only pollster that has been very active in 2014 that we have yet to hear from this summer is Angus Reid Global. As Angus Reid has had the race far closer than anyone else has, it will be interesting to see if they will change their tune or continue to set themselves apart from the pack.