Monday, April 30, 2012

2012 Alberta provincial election analysis

Results map
One week ago, Albertans went to the polls and elected their 12th consecutive Progressive Conservative government in a row. While the Tories have been in power for 41 years now, this result was a massive surprise victory. Every single poll conducted during the campaign pointed to a victory by the right wing Wildrose Party. Some polls suggested they would win by up to ten percentage points. However the election result was the inverse of this. The Tories won the election in a landslide- winning the popular vote by 10%!

So, what happened?

Well, first of all, some blame has to be put on the polling companies. There is no way that going into the last weekend of the campaign the Wildrose Party was ahead by the numbers the polls were showing. You just don't get nearly 20% of voters changing their minds within a few days, especially if no large news story occurs that would derail the campaign. Sure, there were some racist and homophobic comments made by candidates, and sure they weren't condemned by leader Danielle Smith, but was that enough to convince nearly one in five Albetans to change their voting intentions in the last weekend? The comments weren't even made that close to the actual election date. And polls conducted after the comments were made still showed the Wildrose Party ahead.

One thing that may have contributed to a large late swing in voting intentions is the large number of undecideds reported in the last few polls. Clearly these were voters who were lukewarm about voting Tory, and were taking a look at the Wildrose Party but ultimately decided against such a risky vote. These undecideds must have gone en masse to the Tories just before the election.

Another factor that much of the media speculated as to why the Tories won was strategic voting by Liberals and NDPers. While it is true many regular Liberal and NDP voters voted for the Tories in this election, their minds were made up about voting Tory long before the election date. If the polls got one thing right on election day, it was the popular vote percentage for both the NDP and the Liberals. This indicates that there was no last minute swing of Liberal and NDP voters towards the Tories.

In the end, the Tories won 61 seats out of the 87 seats up for grabs. The Wildrose Party still finished second with 17 seats. The Liberals finished third with five, and the NDP won four seats. These seat totals point to a landslide victory, but the popular vote was closer. The Tories won 44% of the vote while the Wildrose Party won 34%.

The Wildrose Party was relegated mostly to rural southern Alberta, which is where they won most of their seats. All but four of their seats won were in rural southern Alberta, and only one of their 17 seats was in northern Alberta. They won just three urban seats; two in Calgary and one in Medicine Hat. This lack of success in northern Albertan and in Calgary, where they were expected to do well, resulted in their loss.

The Tories won the election by winning most of Calgary and Edmonton, as well as nearly sweeping Northern Alberta. They were especially strong in suburban Calgary and Edmonton, where voters are more socially progressive and less likely to support a party like Wildrose, but are still too right wing to vote Liberal or NDP.

The Liberals did much better than most predictors had thought. While their popular vote was around what was expected, their seat total was very efficient. They won five seats thanks to some strong candidates as well as good demographics for the party in urban ridings. The Liberals were still relegated to three seats in Calgary and two in Edmonton.

As expected, the NDP doubled their seat total to four, but were still not able to win any seats outside of Edmonton. They did win less seats and less votes than the Liberals however, and that must be seen as a disappointment for the party.

Where I went wrong.
My prediction of the Alberta election was my worst prediction ever. I will be the first to admit that it was inexcusable. I was the only predictor that got less than 50% of the ridings correct (I got only 45%). This is almost impossible to attain, because in most elections there are more than 50% of seats considered “safe”, meaning you are guaranteed to be at least half right. That wasn't the case however in this election, because of the expected Wildrose surge.

When two parties battle each other for similar voters, as was the case in this election, it can be very difficult to determine where one party will win and lose. Especially, considering a lack of historical precedent to help guide me in making a prediction. I used poll data that projected a near-10 point victory for the Wildrose Party, which resulted in a prediction of 62 seats. However, the reverse happened. The Tories won by 10 points, and they won 61 seats. While I got the parties wrong, I can at least be happy to know I predicted the fact that a 10 point victory for a party would result in a 60 seat landslide. Especially considering these two parties were duking it out for the same right wing voter. If Wildrose had won by a margin that polls were suggesting, I would have probably been very accurate.

I cannot ignore however the fact that I did get less ridings correct than other predictors. So, what did I do differently that would result in that? Most predictors took into consideration one final poll conducted by Forum Research on the day before the election which showed the two parties very close to each other, albeit with Wildrose still ahead. I made my predictions before this poll was released, and therefore did not take it into consideration. However, even if I had seen the poll, I probably would not have put much weight into it. It appeared to be an outlying poll, inconsistent with polls conducted not just by other pollsters, but one from the previous day conducted by the same Forum Research firm. Plus, the poll was conducted on a Sunday, which is probably the least reliable polling day of the week, as people are less likely to do polls on that day. I had also hypothesized that many religious Wildrose supporters were not polled because it was a Sunday, and was one of the reasons why that poll showed a low number for them. But, I was clearly wrong.

I can take solace in a few things, however. Every riding that Wildrose won, I had predicted they would. All but one seat I had predicted the Tories would win, they did. Same for the NDP. Also, I was the only major predictor to have the Liberals winning any seats, let alone the four I predicted. Out of those four, they won three plus two more I had not predicted.

Usually after an election, I analyze each riding that I got wrong. However, because I got most ridings wrong, this would be too lengthy. So, I will take a look at the ten ridings I was off by the most:

10: Dunvegan-Central Peace-Notley
The fact that the Tories even won this riding has to be one of the biggest riding level surprises of the election. Why? Well, this was the second best riding for the Alberta Alliance back in the 2004 election, where they lost the seat by just 300 votes. So, it this was one of the ridings I thought Wildrose would have in the bag. However, it was not to be, as it went the same way as most of the rest of Northern Alberta. This was still a close race, however. The Tories won this seat by 200 votes, an even narrower margin than 2004. I however expected Wildrose to take it easily however, as I underestimated the Tories by 20%, and overestimated Wildrose by 15%

9. West Yellowhead
Some have described West Yellowhead as being the most left wing rural Alberta riding. That is why it wasn't a huge surprise that the Wildrose Party didn't win here. After all, it was one of their worst ridings in in 2008, getting just 4%. However, this was the riding of the Alberta Party leader, Glenn Taylor. I figured much of the left wing vote would go to him, instead of the Tories, allowing Wildrose to come up the middle and win. While Taylor did get 17% of the vote here (more than I predicted, too), the Wildrose Party underperformed my prediction by 11%, and therefore didn't have the votes to come up the middle. The Tory vote held its ground, and they outperformed my prediction by 21%.

8. St. Albert
I got this riding wrong because I thought the Wildrose Party had a strong, credible candidate in city councillor James Burrows. The seat was open, and I thought that would be enough for him to win. However, what I should have taken into consideration is the riding's large left wing base. In 2008, 46% of the riding voted for left wing parties. Neither the Liberals nor the NDP were running credible campaigns here. Where were those votes going to go? Well, they went to the PCs, who won this seat with the same percentage as in 2008. What the Tories lost to the Wildrose here, they gained back from the Liberals and NDP. Burrows underperformed my prediction by 13%, and the Tories overperformed by 21%.

7. Peace River
This riding was a surprise because of the decent showings the Wildrose and the Alberta Alliance have had here in the past. Wildrose got 11% of the vote in 2008, and the combined Alberta Alliance and Social Credit vote in 2004 was 14%. However, this remote riding in Northern Alberta is not very culturally similar to the Wildrose base in rural southern Alberta. Wildrose got just 28% of the vote here, underpolling by projection by 15%. The Tories held their ground, getting 21% more than I had predicted.

6. Calgary-North West
An open seat, in suburban Calgary, with the Wildrose ahead in the polls in the city? Why wouldn't one predict this seat go Wildrose? They did get a strong 15% here in 2008, as well. Maybe because the riding had a large Liberal vote willing to go PC to stop Wildrose. Well, that's what happened. The Liberal vote collapsed here, going from 30% in in 2008 down to 7%. This benefited the Tory candidate who actually increased the popular vote for the PCs from 46% to 52%. This was 22% higher than what I predicted. Wildrose was 11% lower than my prediction.

5. Calgary-Glenmore
This riding was another big surprise on election night. After all, the Wildrose had already won this seat in a by-election in 2009. So, not only did they have an incumbent in former leader Paul Hinman, they also had the benefit of already having won the riding. However, on election night Hinman only increased his portion of the vote by 1% from his by-election victory. However, that was only good enough for 38% of the vote. What did him in however was strategic voting from the Liberals. This riding had a large Liberal based that collapsed into the Tories. The Tories gained 22% from their by-election showing to secure 49% of the vote. That was 22% more than I had thought. Meanwhile, The Wildrose underperformed my prediction by 14%.

4. Calgary-GreenwayWhat really did in the Wildrose Party in this riding was their controversial candidate, Ron Leech. Leech had made comments during the campaign suggesting he had an advantage in the race, because he was white. His opponent was Manmeet Bhullar, a Sikh. In the end it was Bhullar who had the advantage, winning the race easily. The Tories increased their proportion of the vote by 11% from the redistributed 2008 results in this new riding. This came thanks to Liberal voters which made up 30% of the riding in 2008, but which collapsed to 11%. It was difficult for me to figure out how much those racist comments would hurt Leech. In the end, I still predicted he would win, which was a mistake. Bhullar got 22% higher than my prediction, while Leech underperformed by 17%.

3. Red Deer-South
With the Wildrose Party having a large lead in southern Alberta, it was easy to predict that Red Deer-South would be caught up in their surge. However, on election night, urban ridings, even in Southern Alberta stayed with the Tories. This included both Red Deer ridings. Red Deer-South was not caught up in the Wildrose wave, and thus I got this prediction wrong. Wildrose got 35% of the vote, 12% lower than I predicted. The Tories got 44% here, 23% higher than I expected. This was partially thanks to a collapsed Liberal vote, which went down from 27% down to 7%.

2. Calgary-Varsity
This was a Liberal seat that was being vacated by its outgoing incumbent Harry Chase. This left the riding open, and hard to predict. I figured Wildrose would win it, since they were ahead in Calgary. However, those Liberal voters went to the Tories en masse, securing the victory for their candidate, Donna Kennedy-Glans. The Tories got just 37% of the vote here in 2008, but this was increased to 46% on election night. The Liberals still held their own, but their share of the vote shrank from 47% to 21%. Wildrose wasn't even strong enough to win this, getting just 26% of the vote. I thought Tory voters would strategically vote Liberal in this riding, instead of the other way around, which is why I messed up here. I overestimated the Liberals by 5 points, while I underestimated the Tories by 24%! And, I overestimated the Wildrose Party by 15%.

1. Banff-Cochrane
Finally, the riding that I got the most wrong was Banff-Cochrane. This riding was the only rural riding in Southern Alberta to not go Wildrose. I adjusted my numbers in this riding as if it was just like the rest of Southern Alberta, but the riding is very different from the rest of the region. It is more left wing, shall we say, as it includes some resort communities like Banff. However, the riding was an open seat, meaning I thought the Tories wouldn't do very well at all. That's where I made my first mistake. 49% of voters in this riding in 2008 voted for left wing parties. And they weren't going to vote Wildrose. So while the Wildrose candidate managed to get 37% of the vote, many left wing voters held their nose and voted Tory to circumvent this. The Tories won the seat with 42% of the vote, a loss of just 7% of the vote. I had them only getting 15% of the vote (a 27% difference). This was a big error on my part, I'll admit. I was closer with my Wildrose numbers though, just overestimating them by 7%. I also overestimated the Liberals and NDP by 10% and 12% respectively to really blow this one. I am however, comforted by the fact that I did say if the Tories were going to win one riding in southern Alberta, it was going to be this one.

Results analysis
With my failed predictions out of the way, it's time to look at what actually happened to get an insight of party strengths and weaknesses on election night. I have provided maps for each of the four main parties, showing their strengths and their weaknesses.

PC strength

Progressive Conservatives
On election night, the Tories saw their strongest showing in suburban Edmonton and Calgary. If you think about it, these are the types of voters that feel the most at home with the PCs. The are economically conservative, but socially moderate. Too moderate to be attracted to the Wildrose Party. The Tories also did well in Northern Alberta, but won most of their races in that region by close margins. If the WRP wants to win in the future, it will come by winning the north.

The Tories saw their weakest numbers in rural southern Alberta and in the urban cores of Edmonton and Calgary. In rural southern Alberta, they were outflanked on their right by the Wildrose Party. In urban ridings in Calgary they were outflanked on the left by the Liberals, and in Edmonton they were outflanked by the Liberals and the NDP.

Wildrose strength

The Wildrose Party saw their strenghth relegated to rural southern Alberta. A sparse land of cowboys and farmers. This areas is one of the most right wing in the county, and is the most receptive to the Wildrose Party. Outside this rural zone, they won just one northern riding (Lac La Biche-St. Paul Two Hills), two ridings in suburban Calgary, and Medicine Hat. Wildrose still did quite well in Northern Alberta, and in southern Calgary, two areas they will need to actually win seats to win the election.

The worst areas for Wildrose were in Edmonton and in central Calgary. The City of Edmonton was a dead zone for the party. Even though some polls had the party leading in the city, when push came to shove, they were not able to even break 25% in any city riding. In all but a few ridings, the party was did not even break 25% of the vote. We know Wildrose did poorly in 2008 in Edmonton, and despite the polls they did poorly once again. The party's next poorest area was in a handful of central Calgary ridings which were Liberal-PC races. The demographics of those ridings were not conducive to supporting a party like Wildrose.

Liberal strength

2012 was a horrible election for the Liberals, who were lucky to win the five seats that they did. The party won two seats in Edmonton and three in Calgary. Save for a few ridings, most of their strength was concentrated in the two major cities. Edmonton was the best part of the province for the Liberals, despite winning less seats there than in Calgary. There were only five ridings in the city where they did not break 10%. Four of those ridings were won by the NDP. Calgary was the next best region for the Liberals. There were nine ridings in the city where they broke 10%, located mostly in the central urban part of the city, and in the eastern part of the city where there's more immigrants.

The worst area for the Liberals was of course everything else. Outside of Edmonton and Calgary there were only five ridings where the Liberals broke 10%. Three of those ridings were still urban in nature, being in St. Albert, Red Deer and Lethbridge. The other two were located in the western part of the province where there are more left leaning communities like Banff. Rural Southern Alberta was the worst of the worst for the Liberals. In rural ridings south of Red Deer, only in Banff-Cochrane and the semi-urban Cypress-Medicine Hat riding did the Liberals even get more than 3% of the vote.

NDP strength

Much like the Liberals, the City of Edmonton was the best part of the province for the NDP. Unlike the Liberals however, the NDP was dead in the water in Calgary. Edmonton was where the party won all four of its seats. The NDP broke 10% in all but five ridings in Edmonton. All five of those ridings were located in the more suburban southwest corner of the city. Their strongest ridings were of course in the urban part of the city. Outside of Edmonton, the NDP broke 10% in only six ridings, including both Lethbridge ridings. Outside of Edmonton, the best riding for the NDP was Lethbridge-West where their candidate, Shannon Phillips finished second, losing by 1100 votes.

Suburban and exurban Calgary was the worst area in the province for the NDP. The party received less than 5% of the vote in nearly every single one of these ridings. In Calgary, the NDP only broke 6% in three ridings. Rural Alberta was the next worst area for the NDP, especially in Central Alberta. In Central Alberta, the NDP only broke 6% in four ridings.


  1. You are wrong.
    Well about part of it. Polls constantly showed a huge portion of the electorate was undecided. I argue that polls WERE correct, that voters who told them pollsters "I am certain to vote" were, 3 days before the election, 44% wildrose and 34% PC. What happened was the undecided voters voted en masse for the PC Party. I'd say if you took a poll the morning of the election it would have showed wildrose having a 2-5 point lead.

  2. I did not take into consideration the final poll but was more accurate than those who did. Why? Two reasons.

    1 - Uniform Swing
    2 - Ratio Swing

    I've had a discussion with you (the webmaster, for others reading this) about using ratio swing. I know you prefer the Uniform method, but I still emphatically insist that the ratio method is needed.

    I DID apply a uniform swing, making me more accurate than those who used ratio along. Where? Inside and outside the cities. I gave the PC Party a 3 point advantage in Calgary, and 5 points in Edmonton. Wildrose got a 5 point advantage in Rural ridings, but only 3 in the "Towns" like Red Deer, etc.

  3. Dunvegan-Central Peace-Notley
    Sorry for all the comments in a row, but do not feel bad about this. I was WAY off with Westmount in the 2011 Federal election for the same reason. Number based predictions have their weaknesses.

    Also, I find your maps to be very enlightening.

  4. I should have also mentioned the need to factor "momentum" into my numbers; something that has bitten me in the past. But even if I had factored in some pro-PC momentum, it would not have been enough to even be close.