Monday, September 22, 2014

2014 New Brunswick Election - Final Projection

New Brunswickers head to the polls today to elect the 49 members of the New Brunswick Legislative Assembly. Heading into the campaign, it appeared as though the opposition Liberal Party was headed for a certain landslide majority government. However, following a gaffe-filled CBC interview with Liberal leader Brian Gallant, the polls have tightened in the province, and nothing is certain in today's vote.

Polls have been few and far between in this campaign, which has resulted in few projections on my part. In fact, I have only done one other projection in this campaign. This is unfortunate, because it has meant I have not been paying enough attention to the election to come up with a decent model. Nonetheless, I will use my basic vote distribution model for this final projection to try and come up with some sort of numbers. For this final projection, I have used two polls in an attempt to come up with a projected seat count. These two polls are one published by Forum Research last night which shows the Liberals and Tories in a literal tie (40% a piece), and one conducted by Corporate Research Associates last week, which showed the Liberals ahead 45-36.

There has definitely been a shift, in what little polls there have been, from the Liberals to the Tories, at least since Gallant's interview with the CBC on September 12. This means it is quite possible that the momentum will carry forward, for the Progressive Conservatives, into the election today, and win in a close election. How much momentum they still have is something one can only speculate on.

Perhaps including the older CRA poll in my model may skew my projection in favour of the Liberals, when it is actually the Tories that have the momentum. However, the smaller sample size of the poll has ensured its weight is much smaller than in the Forum poll. Nonetheless, it is enough for my model to show the Liberals with a narrow 42%-38% lead. This translates into 30 seats for the Liberals and 19 for the Tories.



Another factor that may help the Liberals win the election is geography. The electoral map of New Brunswick has helped the Liberals out in the past. In the 2006 election, the Tories won the popular vote by a mere 0.4%, but it was the Liberals who won a majority government, winning the election by three seats. However, the map of New Brunswick's ridings has changed since then, and the map may be fairer than the last. In fact, a quick extrapolation of the 2010 results shows both parties would win close to the same number of seats if they were tied in the popular vote. While the current map might have been fair for the 2010 results, the distribution of the votes in this election may be very different. Despite the tie in their poll from last night, Forum Research shows the Liberals ahead in Central and Eastern New Brunswick, while they are far behind in the South. If this poll shows an accurate regional breakdown, then we can assume that much of the Tory support is sitting in the giant vote sink that is Southern New Brunswick, where my model shows them winning all but two seats. However, Southern New Brunswick only has 11 seats, and even if the Tories won all of them, they would be nowhere close to winning a majority.

If the Tories are to win the election, they will need to close the gap in the rest of the province, outside of the South. My model shows quite a few marginal seats that the Progressive Conservatives could win. The closest seats in my model (where the Liberals area ahead) are Memramcook-Tantramar, Moncton South, Shippagan-Lameque-Miscou, Southwest Miramichi-Bay du Vin, Fredericton-York, Oromocto-Lincoln and Saint John Harbour. A near sweep of these ridings could be enough for them to win the election.

Other than the Liberals and Progressive Conservatives, my model shows no other parties winning any seats. Despite so much optimism from the NDP in the lead up to today's vote, the polls have not been favourable to the party (Forum has the NDP at 12%), as voters are shifting back to the more traditional two parties. This is not to say the NDP won't win any seats. There are a number of seats they have targeted, such as their leader's riding of Fredericton West-Hanwell, Saint John Harbour and Independent MLA (running for the NDP) Bev Harrison's riding of Hampton. The populist People's Alliance party (who are only running in 18 districts) are also targeting a number of seats, such as their leader's seat of Fredericton-Grand Lake, and their deputy leader's seat of Southwest Miramichi-Bay du Vin. Their deputy leader actually mentioned to me on Twitter that he was neck and neck in the seat, and that he would send me internal polls to prove it. I didn't receive any, and so my model has him at just 8%. The Greens have been polling well-ish (Forum has them at 6%), but we all know they tend to over-poll. Their best bet is the urban progressive-leaning seat of Fredericton South, where their leader is running. Due to the lack of detailed regional polls in the election, it is impossible for me to accurately gauge how strong local candidacies are of any party, especially the smaller parties, which tend to focus their entire campaigns on a small handful of seats.

Leading mother tongue by census subdivision


My overall projection map shows the geographic polarization of New Brunswick. The Tories are strong in the socially conservative “Baptist Belt” in south and west of the province, while the Liberals are stronger in the Acadian areas, along the Atlantic coast, and in the northeast. This is the general pattern in New Brunswick politics, and the key to victory tends to be in breaking into the other side's traditional territory, usually done with strong local candidates. New Brunswick election maps rarely show this polarization as much as my projection map does, and it is likely that the strength of certain local candidates will make this particular map look quite different from the actual electoral map that will be produced by the voters today. 

New Brunswick religion map by county
 
Projected results by riding (ridings coloured by how they went in 2010, using transposed results): 


Polls close at 8pm (7pm Eastern). 

Thursday, September 4, 2014

2014 New Brunswick Election Projection #1 (Sept 4)

Projection Map #1
We are now two weeks into the New Brunswick provincial election, since the writs dropped August 21, which marked the beginning of the campaign. This means that there are just two and a half weeks to go before New Brunswickers head to the polls on September 22. And, only this week have there been any polls released. Now that there has been some polls, I can finally do my first seat projection of the campaign.

Corporate Research Associates (CRA), Atlantic Canada's main pollster, released a poll on September 2nd showing the Opposition Liberal Party firmly in the lead at 48%. The governing Progressive Conservatives were in second, with 28%, and the NDP was at 17%. Forum Research also released a poll, a few days earlier, showing similar numbers; The Liberals were at 46%, the Tories at 31% and the NDP at 15%. Both of these polls were conducted before Labour Day, so there is the caveat of notoriously unreliable Summer polling, but the numbers are on par with CRA polls from the Spring.

Neither pollster provided for any regional or even linguistic breakdowns in their numbers, which will make individual seat projections a lot more of a crapshoot than in other provinces. For my first projection, I took an average of the two polls (weighted based on sample size), and plugged it into my projection model, which is based on the transposed 2010 election results. (See this post for the calculated transposed results of the 2010 election). I also made some minor tweaks to reflect which candidates will be on the ballot (the nomination period having ended), and I also made adjustments in three ridings based on circumstance:

* Tracadie-Sheila: I reduced the support for the NDP in this riding, because the party saw a larger-than-usual vote share in 2010 because their leader at the time ran in this riding. (I based my math on what the NDP should have received in this riding in 2010, if it saw the same swing as the rest of the province.)
* Fredericton West-Hanwell: Conversely, I raised the support for the NDP in this riding, because their leader is running in it. (I based my math on the increase that NDP leader Dominic Cardy saw when he ran in a by-election two years ago in Rothesay, compared to province-wide polling at the time.)
* Carleton-Victoria: In this riding, the Liberals suspended their candidate (Andrew Harvey) based on fraud charges. Because the nomination period is over, he will remain on the ballot. I have yet to come up with a very good math-based solution to base my projection in this riding, but for now, I weakened Harvey's candidacy based on a similar scenario that occurred in the federal election, where a Liberal candidate appeared on the ballot after losing the party's support due to scandal. Assuming most of his vote will go to the NDP, I increased the NDP's share in this riding accordingly, to compensate. I may want to tweak the numbers in this district in the future, as my model still shows the Liberals in second place.



As always, I will be making further adjustments to my model to reflect candidate strengths, and other factors in the coming weeks. But for now, my model shows a large Liberal majority government. According to my projection, the Liberals would win 38 of the 49 seats in the New Brunswick legislature. The Tories would win 10 seats (almost all of them in the socially conservative “Baptist belt” of southwestern New Brunswick), and the NDP would win one seat (Fredericton West-Hanwell, where Cardy is running). This would be a 25 seat increase for the Liberals, and 32 seat decrease for the Tories from the 2010 election. For the NDP, it would be their first seat won since 2003. For the Liberals, it would be their largest electoral victory since 1995, and it would be the worst defeat for the Tories since then. In that election, the Liberals won 48 of 55 seats, and won 52% of the vote to 31% for the PCs, a similar result to current polling.

Projected results by riding (ridings coloured by how they went in 2010, using transposed results):