Tuesday, May 5, 2015

Final Alberta projection: NDP majority

Rachel Notley's New Democrats are on track to win a majority government in Alberta's provincial election tonight, according to the polls and ThreeHundredEight.com's projection. Brian Jean's Wildrose will likely form the Official Opposition, while Jim Prentice's Progressive Conservatives should finish as the third party in the legislature. Greg Clark's Alberta Party could win one seat, and David Swann's Liberals should be shut out.

This being Alberta, though, we should be ready for anything tonight.

The likely outcome

The New Democrats are projected to win between 48 and 61 seats, putting them comfortably over the 44-seat mark required for a majority government. They are projected to take between 40.9% and 47.6% of the vote. Even at their lowest range, the NDP is projected to win by almost 13 points. The NDP has a higher probability of finishing within the low to average band of seats (48 to 55) than they do in the upper band of 55 to 61 seats. The precise projection awards 55 seats and 44.5% to the New Democrats, which would mark the party's best performance in its history (by far).

Final projection
Wildose is projected to win between 17 and 31 seats and between 24.6% and 28% of the vote. Wildrose has a higher probability of finishing within the average to high band of 25 to 31 seats. The precise projection, of 25 seats and 25.9% of the vote, would be a mixed bag for the party. It would give them the most seats they have won in their short history, but represents a significant drop from Wildrose's 34% from the 2012 election.

The Progressive Conservatives stand to suffer the bulk of the seat losses tonight, and are projected to win between four and 17 seats, with between 22.5% and 26.1% of the vote. There is a much stronger probability that the party will win between six and 17 seats, however, than between four and six. The precise projection of six seats and 23.7% would represent the party's worst performance since 1967, in terms of seats, and 1963, in terms of votes.

The Liberals are projected to win no seats and between 2.9% and 3.4% of the vote. The precise projection of 3.2% represents the party's lowest share since 1982.

The Alberta Party could potentially win one seat. Along with other parties and independents, this group is projected to take between 1.9% and 3.2% of the vote.

Expecting the unexpected

The polls in the final days of the Alberta campaign have been absolutely unanimous, so in normal circumstances we would not be entering into today's vote with much uncertainty. The only real question would be whether Wildrose or the PCs would finish second.

Projection ranges
But because of what happened in 2012, and particularly since this projected outcome is so out of character with the province, we do need to pay attention to the projection's wider bands. These are meant to reflect 95% of potential outcomes, assuming the polls perform as well (or as poorly) as they have in other recent elections.

At these ranges, a series of outcomes are plausible. The New Democrats could win between 31 and 70 seats. That low end still represents their best result in the party's history, and would almost certainly put the NDP in the Official Opposition role (and give them largest opposition contingent Alberta has ever seen).

For Wildrose and the PCs, a victory is still plausible. It requires the polls to be about as wrong as they were in 2012, causing a swing that would still give the NDP the most votes but penalize them in the seat count. The NDP finishing second in the popular vote falls outside of the 95% confidence interval.

Wildrose could win as many as 44 seats, which technically gives them a majority. But that is at the extreme edge of the range, and so extremely unlikely. The PCs top out at 40 seats, meaning a majority is not in the cards for them. Altogether, only the NDP has a realistic chance of winning a majority of seats.

The projection maxes out at one seat for the Liberals, though I suspect they could do better than that (more on this below).

As for the minimum ranges for the PCs and Wildrose, both would be disastrous (one and nine seats, respectively). Both are unlikely to occur in conjunction, though. If the PCs end up that low, Wildrose has likely made some gains. And vice versa.

So, we should not rule out entirely a plurality of seats for either Wildrose or the PCs. But it is unlikely, going by the numbers. There was far more overlap in the projection between the Liberals and the PCs, in Ontario, and the Liberals and the PQ, in Quebec, when both Liberal parties won big majorities in 2014.

Regional breakdown

The New Democrats are projected to finish first by wide margins in both Calgary and Edmonton, while the race is tighter in the rest of the province.

Regional projections
In Calgary, the New Democrats are projected to win between 34.9% and 40.6% of the vote, giving them between 14 and 18 seats. The party has been making gains in the city throughout the campaign, finally moving in front at the end of April. Wildrose is projected to finish second with between 25.7% and 29.3%, winning two to seven seats. While the party is expected to place second in the vote, it will likely finish third in the seats. The PCs are projected to win between four and nine seats, with between 25.2% and 29.1% of the vote. The Liberals should take between 4.6% and 5.4% of the vote, while the Alberta Party could win its one seat here.

Regional tracking
As they have throughout the campaign, the New Democrats are projected to have a massive lead in Edmonton and its surroundings, with between 55.3% and 64.3% of the vote. That should give them between 23 and 28 seats, leaving only zero to four seats for the Tories and one to four seats for Wildrose. The PCs should take between 17.8% and 20.6% of the vote, with Wildrose in third at between 14.4% and 16.4%. The Liberals are projected to take between 3.5% and 4.1% of the vote.

The seat projection in the rest of Alberta has little relation to the popular vote. The NDP is likely to finish first in the region with between 34.2% and 39.8% support, but place second with between 11 and 15 seats. Wildrose should take between 32.1% and 36.5% of the vote, but between 14 and 20 seats. For the Tories, they are squeezed out horribly. They are projected to win between zero and four seats, despite having between 24.2% and 28.0% of the vote. The Liberals, with between 1.1% and 1.3% support, should finish behind the Alberta Party.

The polls

Polling in Alberta has been remarkably consistent. In the final set of polls, which all left the field on or after April 28, the NDP has registered between 37% and 45% support. In fact, in the last five polls - all taken on or after April 29 - the NDP has registered between 42% and 45%. That is a very tight grouping.

Polls making up at least 95% of the projection
In that very last set of polls, Wildrose has been the consensus second place finisher with between 23% and 27% support, compared to a range of between 21% and 23% for the PCs. It is hard to argue with that sort of agreement.

Not that this will stop anyone. There is no arguing with what the polls are saying - the only argument can be with whether or not to believe them at all. There is no reasonable way to look at these polls and conclude that Alberta is a three-way race, or that anybody but the NDP will win. That argument could be made using a whole slew of other factors, but the polls cannot be one of them.

Vote projection over time
The trends are an important reason for that. The New Democrats have only been gaining. Not once since the projection was launched at the start of the campaign has the NDP dropped in support. The NDP started the campaign at 20%, which promptly grew to 25%, then 30%, then 35%, then 40%, and now 45%.

The PCs, by contrast, have been nothing but flat, with a slightly downwards tilt. The party is showing no signs of momentum in the final days. The same goes for Wildrose, which took a hit after the debate and has been stuck in the mid-20s ever since. If there is to be a late swing, it will have to be sudden and not based on any previous movement.

How the leaders fared

For a provincial election so dominated by the leaders, and with a relatively high number of polls, there hasn't been much leadership polling. But what there has been has echoed the voting intentions surveys - great for Notley, middling for Jean, bad for Prentice.

Notley's numbers have been stellar, with about half of Albertans saying their opinion of her has improved throughout the campaign (less than 1-in-10 think otherwise). She has topped all of the polls on who would make the best premier, and her approval rating is through the roof.

Prentice, by contrast, has seen his personal approval ratings plummet. His campaign has had the opposite effect of Notley's, worsening voters' opinion of him in a majority of cases. His approval rating rivals that of Alison Redford before she resigned in disgrace.

For the other leaders, the campaign has not had much effect. Jean and Swann roughly split on approval ratings and whether they have improved or worsened voters' opinion of them. This argues against any sort of advantage for them going into the ballot box.

Riding projections
The most confident prediction I can make is that Rachel Notley will win her seat of Edmonton-Strathcona. In fact, the projection model gives this a 100% probability. Her projected vote share, of between 84% and 98%, is almost certainly exaggerated. The PC candidate, Shelley Wegner, is projected to take 7% to 8% of the vote. I think she'll do better.

Brian Jean will have a closer race in Fort McMurray-Conklin, which he is projected to win with just 56% confidence. His vote share is projected to be between 35% and 40%, with the NDP's Ariana Mancini taking 32% to 37%. The PC incumbent, Don Scott, comes up third with between 25% and 29%. I imagine there will be more of a movement towards Wildrose from anti-PC voters who could instead be wooed by the NDP.

Jim Prentice is expected to win his riding of Calgary-Foothills with 90% confidence, and between 50% to 58% of the vote. Wildrose's Keelan Frey is projected to take between 27% and 31%.

Liberal leader David Swann could be in tough in Calgary-Mountain View, where the NDP's Marc Chikinda is favoured by the projection with 37% to 42% of the vote. Swann is awarded 24% to 28%. I think it will be closer than that, and Swann could defy the model (Liberals have done that before in Alberta).

The most interesting race may be in Calgary-Elbow, where Alberta Party leader Greg Clark is narrowly favoured with between 20% and 33% of the vote. PC incumbent Gordon Dirks is expected to take between 26% and 30%. The projection is a literal toss-up. The wildcard will be Wildrose and NDP supporters.

A note on the Liberals

One major assumption the model has made in this campaign is that the polls will over-estimate the Liberals by about the proportion of candidates the party is not running in this race. As a result, the projection drops the Liberals' vote considerably, as the party is running less than a two-thirds slate. But the latest round of polls have had the Liberals very low already. Perhaps Albertans are more aware of whether or not there is a Liberal candidate in their riding than other voters have proven in other elections where a party did not run a full slate.

If I did not penalize the Liberals as I have done, they would be projected to take 4.9% of the vote province wide (6.1% in Calgary, 4.6% in Edmonton, and 3.4% in the rest of Alberta). This would award them between zero and two seats, with the maximum range going up to three.

I have also done the same thing with the Alberta Party, which did not run even half a slate of candidates. The party is expected to take around 1.5% of the vote. If, instead, I relied on the polls, I'd give the Alberta Party and the other parties a total of 5% support, reducing (in conjunction with removing the penalty from the Liberals) the NDP to 42.7%, Wildrose to 24.9%, and the PCs to 22.7%. The seat projection would be virtually unchanged.

2015 not like 2012 or 2013, but could have its own surprises

The PC dynasty about to meet its maker? Sounds like 2012 in Alberta. The NDP projected to win in a western province? Sounds like 2013 in British Columbia.

Oh, did you not know the polls missed those two elections? I'm reminded every day - apparently I keep forgetting!

There are a lot of reasons that this election is nothing like those two examples.

In the 2012 Alberta election, Wildrose held a lead of eight points over the Progressive Conservatives, on average. In this election, the NDP's lead has averaged 17 points over Wildrose, and 18 points over the PCs. Obviously, that is quite different. The degree of error, or the size of a late swing, needs to be at least twice as large as it was in 2012, and that election was exceptional enough.

But if there was a late swing in 2012, there were few pollsters to capture it. In an election that occurred on a Monday, all but two pollsters dropped out of the field on Thursday or earlier. This time, however, the last three days of the campaign had three pollsters in the field (four polls in total), and the numbers held steady. In 2012, by contrast, the final Forum poll showed the gap between Wildrose and the PCs shrinking to just two points - a hint that something big was about to happen?

No such hint has been given this time.

The dynamics of the campaign were different as well. Anecdotally, there seems to be a lot more anger and fed-uppedness with the Prentice Tories than there was with the Redford Tories. There seems to be more genuine interest and energy surrounding the NDP campaign than the Wildrose campaign in 2012. The last week of the Wildrose campaign in 2012 was a very bad one. This time, the only party that had a particularly bad last week was the PCs.

Rachel Notley is polling better than Danielle Smith was, and Jim Prentice is polling worse than Alison Redford was. There are three parties garnering a lot of support, giving vote-switchers from these three parties two viable alternatives. In 2012, there was only one alternative.

The only commonalities between the 2012 and 2015 elections are that they are elections in Alberta, and the PCs are trailing. That's it.

In the 2013 B.C. election, the New Democrats held a lead of only eight points over the B.C. Liberals, again half the size of the lead the Alberta NDP currently enjoys. And the B.C. Liberals had spent the entire campaign closing the gap. Here, the PCs have spent the entire campaign widening it.

In B.C., Christy Clark had moved neck-and-neck with Adrian Dix on who would make the best premier, and her party was polling better than the NDP on the economy. Now in Alberta, Notley is polling about twice as high as Prentice on the premier question, and is beating him on the economy.

Notley's polling numbers are far better than Dix's were. Clark's were also better than Prentice's.

And with the NDP ahead, many voters seemed comfortable voting for the Green Party. An exit poll showed that British Columbians expected the NDP to win. In Alberta, no party is filling that equivalent spoiler role.

But there are factors related to this election that could make the results tonight different from what the polls are suggesting.

The most important one is history. The Progressive Conservatives have been in power for almost 44 years, and change is a scary thing. A lot of voters may finally balk in the ballot box, with both Wildrose and NDP voters flipping back to the PCs. The polls give no indication that something like this will happen (in some polls, Wildrosers choose the NDP as their second choice, and New Democrats choose Wildrose), but theoretically it could. A lot of Albertans with infinitely more knowledge of the local situation than I believe it will happen - but that opinion cannot be based on any of the data I am seeing.

The biggest danger for a surprise result may be in the seat count. Things are still close enough in Calgary that a marginal swing could flip a lot of seats. Ditto for the rural parts of the province. Give the PCs seven-points' worth of NDP support, and you have a PC plurality. Give them even less, but a sprinkling of Wildrose support, and you have the same result.

So hold on to your hats. The polls are unanimous. In any other jurisdiction we'd all be very confident in the result. But this is Alberta, and not only is the polling history mixed but decades upon decades of experience are arguing against this surprising result. If you listen to your gut, to your anecdote, to conventional wisdom and assumptions, maybe you see the PCs pulling it off. If you want to go by objective, cold, unapologetic numbers, though, the NDP has it.

If I was a betting man, I'd go with the numbers. But then again, my gut has been feeling uneasy all week.