Monday, February 6, 2012

Redford's Tories on track for landslide in first projection

Premier Alison Redford is expected to present her first budget to the Alberta legislature this week, the first step towards the province’s next election scheduled to take place within the next four months. But if an election were held today, Ms. Redford’s Progressive Conservatives would win a landslide victory.

The latest seat and vote projections based on all publicly released polls indicate that the Progressive Conservatives have the support of 45.3 per cent of Albertans, down 7.4 points since the 2008 provincial election. But faced with a divided opposition, the Tories are projected to win 73 seats in the 87-seat legislature, six more than they currently hold.

You can read the rest of the article on ThreeHundredEight's first Alberta projection at The Globe and Mail website here.

With the Alberta legislature returning to work and an election call coming in four to six weeks, ThreeHundredEight is now launching its vote and seat projection model for the province. It's a little different from the models I've used in the past.

The graphic on the top of the page shows the projected vote and seats for each of the parties in Alberta. It also shows their low and high ranges. This is one of the changes that ThreeHundredEight will be making going forward, and that is a measure of the uncertainty involved in polling and projections.

This uncertainty is measured by the volatility in the polling. For example, if recent polls have one party's support varying by no more than three points, the volatility for that party will be low. In this Alberta projection, this applies to the Liberals and the New Democrats. On the other hand, if polls vary by a lot more points then the volatility will be quite high. This is the case for the Progressive Conservatives, projected to have the support of between 37.8% and 52.8%, and Wildrose (16.7% to 29.7%).

The projected result, however, is the most likely. But the projected vote is not just the result of an aggregation of the polls. Though weighing the polls by date, size, and polling firm accuracy is still most important, the polls are also adjusted according to how they should be expected to diverge from actual voting behaviour. After analyzing all recent provincial and federal elections where detailed polling data was available, I've found that the number of seats each party has in the legislature is the best predictor of how voting intentions differentiate from voting behaviour.

The first and second largest parties tend to out-perform the polls, perhaps because of their better organization, higher fundraising, and better chances at forming government. Third and fourth parties tend to under-achieve their poll results, likely because of their weaker organizations and susceptibility to strategic voting. In these cases, the difference is roughly by a factor of 5% or so (not percentage points). If a party is not represented in the legislature, they tend to under-perform the polls by about 25%.

This model, unlike previous ones, is a regional model. I have split Alberta into three regions: the CMAs of Edmonton and Calgary and the rest of Alberta. Aside from Forum Research, every active pollster in Alberta splits up the province in this way.

The seat projection is done at the regional level, and here again volatility comes into play. This determines the low and high ranges of the projections in each and every riding, and from these vote ranges the seat ranges are determined. If a party's high range is greater than the low range of the party projected to win the seat, that indicates that the seat is at play and this is how the seat ranges are determined. When there is less volatility in the polling, the ranges shrink. That is why the ranges for the NDP and the Liberals are so small, while that of the Tories and Wildrose are so large. Their polling has been all over the place.

The image at the top of the right-hand column shows the regional vote and seat projections. It can be clicked on to see the regional vote and seat ranges. The riding projections can be accessed by clicking on the "Alberta Riding Projections" banner under the main projection image at the top of the page.

Finally, the chart at the very bottom of this page shows which polls are included in the projection and what weight the polls of each firm are given. There are always older polls included in the projection than those showed, but their weights are too small to register. Also note that the weight chart lumps multiple polls for each pollster together. The main reason why Forum's polls are weighted more heavily than Léger's is because there are two of them to Léger's one.

As the projection is updated, I will put up the tracking charts that will follow the shifts in the vote projection and vote ranges over time. It should give a good road map to the campaign as it unfolds.

The potential for a good contests exists, as the vote ranges show the margin between the Tories and Wildrose could shrink to as little as eight points, and Wildrose could win as many as 24 seats. Those are, at this stage, unlikely outcomes. But hopefully we'll see plenty of action in the coming weeks both in the polls and on the campaign trail.

26 comments:

  1. The NDP has not measured below 13% in an opinion poll in over a year according to wikipedia (http://en.wikipedia.org/wiki/28th_Alberta_general_election#Opinion_polls ) Yet their high range is 12.7%??

    ReplyDelete
  2. Please read the 7th and 8th paragraphs to find out why.

    ReplyDelete
  3. I understand the methodology. I just think it's nonsense.

    In the run up to previous elections, it was the small parties like the NDP that polled poorly. Before the 2008 writ, I recall a couple of polls that had the ndp in the 4-5% range, and they ended up at 8.5%

    ReplyDelete
  4. In the last three polls of the campaign in 2008, the NDP averaged 9.3%. In fact, during the entire campaign only two polls out of nine had them below 8.5%. The others ranged from 9% to 14%.

    We're talking about 0.3 points in this case, however. For all we know, the three recent polls had the NDP at 12.5% to 12.7% and just rounded up.

    ReplyDelete
  5. Sorry, two polls out of eight. None had them lower than 7%, though, at least according to a quick look at Wikipedia's page for that election.

    ReplyDelete
  6. Glad to see the Alberta projections beginning! Out of curiousity, how did you derive the Evergreen Party numbers? I don't think they've been included in the recent polls; did you use the "Other" results from those polls to estimate?

    Also, are you making any adjustments for "star" candidates, or where the leaders are running? For example, Danielle Smith in Highwood would probably outperform the WR poll numbers in the area; Liberal leader Raj Sherman in Edmonton-Meadowlark might do better than the Liberals will in general; and Glenn Taylor, the Alberta Party leader, who was the very popular mayor of Hinton, is running in West Yellowhead which would probably help him outperform the AP poll numbers in the rest of rural Alberta.

    I can think of a few other examples too, but they might be harder to handicap. E.g. some candidates who are well-known in the area, or who have been MLAs before, etc. But you may be wanting to go with raw numbers only, rather than trying to bring more subjective assessments in, which I completely understand. Or it may be just too early for that!

    ReplyDelete
  7. Brandon,

    For Evergreen, I subtracted a little from the "Other" results to represent independents and fringe parties, and then gave the rest to Evergreen. Without a seat in the legislature, they were reduced further.

    I've made some adjustments for "star" candidates like Smith and Taylor, but nothing will be finalized (including whether the smaller parties will run full slates and the role of independents) until the official candidate lists are released mid-campaign.

    Hopefully a riding poll or two in some of these interesting ridings might appear. Does anyone remember if riding polls were released in 2008?

    ReplyDelete
  8. I don't recall prior riding polls, but I would have expected them. Since the overall result of Alberta's elections generally aren't in doubt, individual ridings are the only places pollsters could find news.

    After all, there are only so many ways to write an article that says Ralph Klein is going to win by 30 points.

    ReplyDelete
  9. 84% of the seats with 45.3% of the vote. And people still oppose electoral reform?

    ReplyDelete
  10. I absolutely oppose electoral reform. And I think people who support it are working from a misunderstanding of what elections are for.

    The goal of an election is not to produce a government that mirrors public sentiment. The goal of an election is to produce an effective government. Majorities are effective. Minorities are not.

    ReplyDelete
    Replies
    1. You could use the same argument that dictatorships are even more effective. The goal of an election is for the will of the people to be reflected in government. For that to happen electoral reform is necessary

      Delete
    2. The will of the people is reflected in the government at the micro level. Each community or neighbourhood is represented by someone selected by a plurality of voters.

      How would you suggest maintaining that level of community representation? I'm open to suggestions for electoral reform, but I'm pretty happy with the system we have now so I'm not willing to accept the loss of any of our system's positive features. Community representation is one of those features.

      Delete
    3. There are very stable and effective governments around the world that use proportional representation.

      Delete
  11. "84% of the seats with 45.3% of the vote. And people still oppose electoral reform?"

    Yes is the short answer.

    The long answer is that your approach is flawed.

    There are 84 individual seats with 84 individual vote totals.

    Averaging the vote totals across 84 seperate races creates an artificial and meaningless metric.

    To then claim a disparity between the percentage of seats won and this artificial metric is intellectually dishonest because its an apples to oranges comparison.

    The incorrect assumption that is operative here is that seats won should be close to the average vote.

    What would be more relevent is how close to 50% was the victor in each of the 84 indiidual ridings.

    It is very possible that a party might win most of their seats with a majority or strong plurality (say above 40%) and then do very, very poorly in all others seats.

    This could be common in situations with strong ideological polarization or urban vs rural divides.

    Not saying that is happening here.

    But such a hypothetical and very democratic election result would be masked by taking the seats vs average vote totals approach.

    Population differences in individual seats, which way votes are being split, and the overall efficiency of each party's vote make the average vote metric rather meaningless.

    Its only use is as an argument for electoral reform.

    Which is all well and good but it would be helpful if proponents of reform would make their arguments without diminishing the collective intelligence of Canadians by using faulty stats.

    ReplyDelete
  12. Ira,
    Elections can be for many purpose, I think the creation of an effective government or the mirroring of public opinions are both legitimate reasons.
    As for effective government, I don't think majorities hold a monopoly over this matter. Germany also employ an parliamentary system, & they never had a early election since WWII. Most of their governments are coalitions, with instances including the two biggest parties! (Imagine a Liberal-Conservative coalition). Germany's government has been relatively efficient IMHO. Of course there are examples of Proportional Representative elections producing strings of unstable government & maybe therefore ineffective (Israel & Italy) but there are a lot of examples to the contrary as well.

    ReplyDelete
    Replies
    1. The difference with Italy and Israel is that they have a pure PR list instead of MMP which acts as a more stabilizing system than PR however still better reflects voting sentiment

      Delete
    2. German's are a remarkably sensible people. In the recent period of free credit for everyone that caused the housing bubble in a great many countries (the US, Ireland, annd Iceland being perhaps the best examples), Germany also had free credit. And Germans didn't use it, becaise they didn't think it was a sensible thing to do.

      So, yes, Germans have made the system work, but I doubt many others could.

      Also, one of the ways Germany does that is by having two of its major parties not run against each other. The Christian Democratic Union and Christian Social Union operate separately in separate areas. The CSU runs candidates only in Bavaria, and the CDU runs candidates only outside Bavaria.

      Imagine if, during the period of one of our big regional parties (the Bloc or the Reform Party), one of the traditional parties had agreed not to run candidates in that party's region, but then to cooperate in parliament. A Conservative-Bloc alliance would have won a majorty in 2008, and come just a few seats short of a majority in 2006 (and election which, under our current system, the Liberals actually won).

      The German political landscape is not relevantly similar to ours. Your comparison is specious at best.

      Delete
    3. Sorry, 2004. The CPC-Bloc alliance would have won majorities in 2006 and 2008, with a near majority in 2004.

      Delete
  13. Crothers, what you say about an artificial metric in only true for some proposed models. If the way proportionality is achieved is on a province-wide or nation-wide basis, then yes, that is problematic. But if it is achieved on a regional or multi-member district basis, that problem is alleviated .

    I'm going to use an Ontario example, since that is what I am most familiar with. Suppose we were operating with regional proportional representation. The region in question is the former city of Toronto, with eleven seats: Beaches-East York, Davenport, Don Valley East, Don Valley West, Eglinton-Lawrence, Parkdale-High Park, St. Paul's, Toronto Centre, Toronto Danforth, Trinity-Spadina, and York South-Weston. The old city of Toronto is fairly politically homogenous, as you can see from looking at the results in these ridings in the last federal election as well as how the municipal wards within these ridings voted in the 2010 municipal elections.

    These eleven ridings could be treated as a region, and have the seats within the region assigned by the percentage of the votes within the region. In those 11 seats, the NDP got 36.3% of the vote, the LPC got 33.3% of the vote, the CPC got 26.3% of the vote and the GPC got 4.1% of the vote. In terms of seats, what that should have yielded is 4 seats for the NDP, 4 seats for the Liberals, 3 seats for the CPC and 0 seats for the Greens. What we got instead was 6 seats for the NDP, 3 seats for the CPC and 2 seats for the LPC.

    While as a New Democrat I am happy to see the NDP get 6 seats, that doesn't represent how the people of the region clearly felt or voted. In a situation in which the city was represented by a region, rather than broken into 11 idiosyncratically divided sub-regions, we would have a much more accurate result to the will of the people without the problem of that artificial metric.

    I would also note from my spreadsheet of the results from those eleven ridings that only three were won with more than 50% of the vote (Davenport, Toronto-Danforth and Trinity-Spadina). Most of the rest were won with support between 40.1% and 43.1%, with one low outlier (Don Valley East at 37.0%) and two high outliers (Eglinton-Lawrence at 46.8% and Parkdale-High Park at 47.7%).

    A regional or multi-member proportional system can produce very locally accurate results while still ensuring that the result of the election is an approximately accurate representation of the will of the electorate as a whole.

    ReplyDelete
  14. Crap, you beat me to it. I was going to do something fancy with my projections as well.

    Oh well, looking good Eric!

    ReplyDelete
  15. Ira - It's a fair point. I've got some stats to support my argument which I'll dredge up when I get home. :)

    Crothers - "But such a hypothetical and very democratic election result would be masked by taking the seats vs average vote totals approach."

    I'm sorry, but the actual popular vote does matter. I'm sympathetic to the parliamentary principle that we elect people, not parties - but at the same time, if you see perspectives go completely unrepresented or massively unrepresented, that's bad for democracy and public discourse.

    As you said, there are 84 for individual ridings, but who decreed that you can only have one member per riding? Ireland, Sweden, and many other countries have regional ridings with multiple MPs. In the last Irish election, the Galway West riding elected 2 MPs from Fine Gael, 1 from Fianna Fail, 1 from Labour and 1 independent (http://electionsireland.org/result.cfm?election=2011&cons=129). They didn't just award 100% of the seats based on 50% (or less) of the vote.

    To say the popular vote is a faulty stat though... wow.

    ReplyDelete
  16. Bah TS - you said it better than I could lol.

    And yah I second that Volkov. Really impressed with the new projection Eric.

    ReplyDelete
  17. Ira - to had to what the others have said, I'd just specify that I'm supportive of the "open-list*" systems typical of northern Europe - not that "closed-list" systems use in southern Europe and Israel. Any system where you don't get to pick who makes up the caucus of each party is a bad one imho, and a step backwards from what we have right now.

    That being said, the track record of open-list systems is pretty good. There are 7 democracies with a higher (nominal) GDP to capita ratio than Canada. All 7 of them use an open-list electoral system (though Australia only does. 6 of them have lower income inequality than us - with Australia roughly on par with Canada.

    Around the world IMHO there's a pretty convincing trend of countries with open-list PR electoral systems outperforming their peers as well. Costa Rica is outperforming its Central American peers. Brazil is outperforming the other BRIC countries. Chile is outperforming its Latin American peers.

    At the same time, did successive majority governments serve the people of Greece well? I'm not going to go out and totally bash first past the post. When Canada was founded, the Westminster system was the best political system in the world. Since 1867 there has been substantial progress though, and I think we should learn from the improvements other countries have made in the mean time.

    Anyways, I'd love to hear what you think on that. And Eric - sorry for threadjacking you here.

    * I'm counting open-list MMP (as in Austria) and STV (as in Ireland, Malta, Australia) as open-list systems, with closed-list MMP as in Germany somewhere between a fully open list system and a closed list system.

    ReplyDelete
  18. Ryan pardon me, I see you're not tackling this from a question of democratic legitimacy vis a vis winning a seat with a majority or clear plurality.

    You're expressing a view that we design a system around getting every possible view point into our house of commons.

    That is the only way in which the popular vote, taken out of context and existing in the abstract, could ever have any actual meaning as a statistic.

    I don't think the only objection to such a system is the idea of communities of interest choosing one representative to represent them as a whole in Ottawa.

    That is to say its OK that a Conservative win a riding with Liberal or NDP supporters because they are advocating for the well being of the community as a whole.

    I think there is another very serious criticism to your idea.

    That is the European example of xenophobic, racist parties winning seats. Especially so in the European parliament.

    Racism and other odious ideologies will probably always exist amongst 10-15% of the population. But never again will it be a majority or plurality in a community.

    I kind of like a system that excludes nutty ideas only held by such a small number of people.

    That's a feature, not a bug, of first past the post in my view.

    ReplyDelete
  19. "You're expressing a view that we design a system around getting every possible view point into our house of commons."

    No. I'm saying that the number of representatives a point of view has in our House of Commons should be proportional to the number of citizens who hold that view.

    "That is the European example of xenophobic, racist parties winning seats. Especially so in the European parliament."

    That's a fair concern, but have you been following American politics lately? I'm afraid first past the post isn't much of a barrier to extremism there these days. If 30% of your voters support neo-fascist ideas, you're going to have problems. Just take a look at some of the statements our PM Mackenzie King made about "the Jews" in the 1930s.

    At least under a proportional system though, I can guarantee that 30% of the vote isn't enough to take a majority of the seats and form government. You can't make the same guarantee under first past the post.

    That being said, I do think certain systems are more resilient to extremism than others. STV and MMP in particular have proven very resilient to extremism. None of the countries using those systems have fascists in their parliaments right now, nor have they ever. Both of these systems foster consensus and push parties to work together and move towards the centre. Under FPTP the incentive is to destroy the parties most similar to yours, and that can lead to some pretty perverse results.

    "Ryan pardon me, I see you're not tackling this from a question of democratic legitimacy vis a vis winning a seat with a majority or clear plurality."

    You see wrong. I partially addressed it, and TS addressed it far better. Take a look at his comments again. Otherwise, I'd suggest familiarizing yourself with how STV and MMP work - as perhaps that is where the misunderstanding lies.

    I don't see how you've addressed my point re: governments that don't have a democratic mandate. Take a look at the BC Election in 1996 (http://en.wikipedia.org/wiki/British_Columbia_general_election,_1996). The second place party took a majority of the seats and governed alone. I'm sorry, but that's not democratic. Period.

    The blowback from that in 2001 (http://en.wikipedia.org/wiki/British_Columbia_general_election,_2001) didn't produce better results either. The government there did have a majority mandate, but the 43% of people who votes for other parties were left with only 2 representatives out of 79. I'm a card carrying BC Liberal, so obviously I was happy with that election result, but even with my preferred party in power I don't think that having such a weak and underrepresented opposition was good for the governance of the province or society at large.

    Anyways, cheers and best regards. :)

    http://en.wikipedia.org/wiki/British_Columbia_general_election,_1996

    ReplyDelete

COMMENT MODERATION POLICY - Please be respectful when commenting. If choosing to remain anonymous, please sign your comment with some sort of pseudonym to avoid confusion. Please do not use any derogatory terms for fellow commenters, parties, or politicians. Inflammatory and overly partisan comments will not be posted. PLEASE KEEP DISCUSSION ON TOPIC.