One week from the election in Alberta, NDP favorite but Wildrose and PC still in the race

My first Alberta projections, ever. I was fortunate enough not to make a call in 2012. I say fortunate because we all remember how wrong the polls were. And this is very much the elephant in the room this year with the crazy polls showing the PC third and the NDP (the NDP!) first: will the polls be wrong again?

Throughout the 2012 campaign, the Wildrose was ahead by a good margin and was heading to a majority government. But on election day, the PC somehow managed to not only win, but did so comfortably. To this day, the 2012 Alberta election remains one of the most blatant example of polls getting it completely wrong. And we still don't really know what happened. The consensus (cop out, depending on your point of view) is a late swing from the Wildrose to the PC. And by late, I mean really late, like 5-minutes-before-voting kind of late. All these conjectures are mostly based on two things: 1) polling firms can be really stubborn and will refuse to acknowledge any wrong doing (some like Ekos will openly admit they were wrong but they are the exception). So they weren't "wrong", people changed their mind. 2) A late poll by Forum that had a late momentum for the PC and a decline of the Wildrose. Except that Forum, while a good firm in general, does tend to have a lot of variations between its polls and I'm not convinced the last poll was really picking up a trend.

Anyway, the 2015 has been even crazier than the 2012 one. Polls show a NDP surge with the incumbent PC in free fall (the type of swings you only see in Quebec). The Wildrose is right there  (despite having lost its leaders and many Mps as recently as December) while the Liberals are, for all intents and purposes, non-existant. Except that obviously I wouldn't blame you if you don't believe these numbers. Beyond the fact that polls have been quite wrong at times (Alberta, BC and Qc 2012), we haven't had that many polls in Alberta. And most are from firms that aren't very well known. You have Mainstreet Technologies that uses the cheap IVR technology and has been polling quite a lot. But they have more than 20% of undecided. We only got one Forum poll in the last 2 weeks and there is this "firm", called 1abvote, that uses Google Survey to measure voting intentions. If that sounds BS to you, they at least provide at lot of details and information. And such a methodology has done quite well in the US presidential election in 2012. Still, it'd be nice to have polls conducted by Angus-Reid, Léger, Ipsos or Ekos. I guess it'll happen towards the very end of this election.

For my projections, I modify the polls slightly. In particular, I don't allocate the undecided proportionally. I think this method tends to overestimate smaller parties. And in the case of Alberta, past elections show that polls have a tendency to underestimate the PC (and not only in 2012). So I decided to allocate more of these undecided to the PC (you can also see that as an incumbency effect). The result is a relatively close race in terms of votes with the NDP ahead by about 6 points. Now, such a lead is significant and in normal times, there shouldn't be much uncertainty. Except that I've learned to be a lot more careful with these numbers. In terms of seats, the NDP is also ahead while the Wildrose is second. It seems vote efficiency isn't the strong suit of the PC, at least not when this party is so low. Notice however that in the Mainstreet polls, the Wildrose has been first, not the NDP. But this one had a really big lead in the one Forum poll. Again, more polls from more firms would be welcome.

Voting intentions; Seat projections with 95% confidence intervals; Chances of winning the most seats

The riding by riding projections are here, along with probabilities of winning in every riding. Remember that the probabilities are estimated through the use of 5000 simulations that account for the uncertainty of the polls as well as of the distribution of the vote/electoral system. Given the high level of uncertainty for this election, I'd recommend looking at the confidence intervals and/or probabilities more than the actual numbers.

The NDP will clean the house in Edmonton but most polls show that the Wildrose is ahead in the rest of Alberta. A majority is also currently unlikely (62% chances of a minority). While the PC is still in the race, its chances are low and a majority seems out of reach.

I don't follow Alberta politics as closely as the federal or Quebec ones, so I have to openly admit that it's hard for me to judge whether these numbers will hold or not (as opposed to 2011 for instance where I was fairly confident that the NDP surge in Quebec was real). I also spent less time building the model than usual. So please remember that when reading these projections or using the simulator.

I'll post final projections before the election. In the mean time, just use the simulator if you see new polls or numbers.