Thursday, April 14, 2011

Liberals gain in Ontario and Quebec

With the two debates now completed, we're entering the final 19 frantic days of the campaign. Three of the polls added to the projection this morning, from COMPAS, Innovative Research, and EKOS, are all pre-debate polls, while only one-third of today's missive from Nanos was taken after the English language debate.

However, I think last night's French debate has a greater chance of having an influence on the campaign. I was expecting a big performance by Jack Layton, as the newfound support the party has in Quebec required some buttressing in the debate. While Layton performed well enough, I don't think he did anything to push his party into second place on election night in the province, and I think we might even see his numbers drop.

Stephen Harper was not very present and will likely not see any boost in Quebec, but both Gilles Duceppe and Michael Ignatieff had a good showing in the debate. Ignatieff performed better in French than he did in English, and was very strong when the debate turned to Quebec's place in Canada. My own perception is that he might have done enough to put himself head and shoulders over Layton and Harper as the federalist option.

Duceppe started off slowly but also woke up at this point of the debate. While he didn't have a home run performance, I think he shored up his own support and we may see his numbers return to the usual 38-40% that the party had enjoyed for the months running up to the campaign.

But in the meantime, the four polls in this projection update (in addition to Nanos's poll from April 8-10 which has been returned to the model) have caused things to swing.
The Conservatives have dropped 0.1 point to 38.9% and one seat to 151, while the Liberals are up 0.1 point to 28.2% and two seats to 75.

The New Democrats are steady at 33 seats, and have picked up 0.1 point to reach 16.8% in the projection. The Bloc Québécois is down 0.3 points to 8.8% nationally and one seat to 49, while the Greens are unchanged at 6% support.

For the most part, the Conservatives have held firm. They gained 0.1 and 0.2 points in the Prairies and Quebec, respectively, while dropping 0.1 and 0.2 points in Ontario and Atlantic Canada.

More damaging is the drop of 0.4 points to 40.9% in British Columbia, especially considering the gain of 0.4 points and 0.2 points by the New Democrats and Liberals.

The Liberals did not see their support shift very much out west, but it did shift out east. The Liberals have gained 0.4 points in Ontario and now trail the Conservatives 41.9% to 35.4%, a gap of 6.5 points. In Quebec, the party is up 0.4 points to 20.7%, while the Liberals have regained the lead in Atlantic Canada with a 0.3 point gain there.

The New Democrats had a good gain in BC and another in Quebec, but the drop in Ontario (though small) is unfortunate for them.

And the Bloc Québécois had a very bad movement, dropping 0.6 points to only 36.7% in Quebec.

The two seats changes have come in Ontario and Quebec. The Liberals have retaken Brampton - Springdale from the Conservatives, meaning incumbent Ruby Dhalla is the favourite once again. In Quebec, the Liberals have also retaken Brossard - La Prairie, held by Alexandra Mendes when the government fell. These are both Liberal seats that are returning to the party in the projection. The Liberals are two seats away from starting to make actual gains.

That will likely take place in Ontario, though a few seats in Atlantic Canada and Quebec are also on the bubble. The Liberals are within 1.1 points of the Conservatives in Ajax - Pickering, Kitchener - Waterloo, and Vaughan, the last riding looking all the more likely to return to the Liberal fold after yesterday's events there.

But several other ridings are also on the tipping point. The Liberals trail the New Democrats by 2.5 points or less in Trinity - Spadina and Sudbury, while they are within three points of the Conservatives in Mississauga - Erindale and Kitchener Centre.

Generally speaking, the projection is behind emerging trends. So while we can't know whether the Liberals will continue to grow their support in the polls, we have a good idea that they will grow in the projection as older polls are dropped off and the more recent reality (putting the Liberals neck-and-neck with the Tories) takes more place in the model. The great thing about the projection, though, is that if the recent Liberal "surge" in the province dissipates, the projection will almost immediately stop the Liberals from moving forward. It treats gains with a raised eyebrow until they can be proved by consistent results.

Note that the riding projections have now been updated to reflect all the independent and minor party candidates. Nominations have closed and Elections Canada has the final tallies of candidates, so the projection is now up to date on how the "Other" vote will likely turn.

For the smaller parties, they have each been given the average result the party managed in the last election in ridings in which they presented candidates. So, for example, the Marxist-Leninists (the party with the most candidates outside of the major five, believe it or not!) receive 0.3% of the vote in every riding in which they have a candidate. That can go up or down depending on how much "room" the other parties provide. The amount of support assumed to go to each minor party in the ridings in which they have a candidate is as follows:

Progressive Canadian 1.2%, Christian Heritage 1.0%, First Peoples 1.0%, Rhinoceros 0.7%, Canadian Action 0.4%, Animal Alliance Environment Voters 0.3%, Western Block 0.3%, Marxist-Leninist 0.3%, Communist 0.2%, Libertarian 0.1%, Marijuana 0.1%, Pirate 0.1%, United 0.1%.

Independents are given 0.7%. When more than one minor party or independent candidate is present in a riding, their results are combined. So, for example, in South Surrey - White Rock - Cloverdale the Other vote is currently 4%, divided between a Progressive Canadian candidate, a Christian Heritage candidate, and three independents. The Others in this riding are assigned 4.3% of the vote, but because of the space available in the riding projection their vote has been reduced to 4%.


  1. I wonder what your thoughts are on how much the Liberal gain in Ontario is actually people changing their minds, and how much is based on realizing who the incumbent is.

    What I mean is, Ontario has been hit pretty hard by the recession, which should lead to anti-incumbent feelings. Is it possible that in pre-election polls, when people aren't paying attention, they said they would vote Conservative because they thought they were opposing Dalton McGuinty.

    Now that the campaign has started and they see the leaders on TV every night, they have realized that the incumbent is Harper, not McGuinty, and have therefore shifted to supporting the Liberals.

    In this sense, the Conservatives never really had a big lead, but rather Ontario has simply been anti-incumbent all along, and since the election is federal and not provincial, voters are focussing their displeasure on Harper.

  2. Éric;

    Error on the riding list; you eliminated the Green from Etobicoke Centre instead of Etobicoke North.

  3. Eric

    Is there some reason that you have not brought the Ipsos flash polls on debate performance into the discussion? You make qualitative observations on the leaders' debate performances, but I would be interested in your take on the numbers.

  4. William,

    Thanks. I'll fix it.


    That could be a factor.


    I haven't purposefully avoided bringing them up, I'm just giving my own view on how I perceived the debates. I'm not sure how valid those polls are anyway, since it can take days for this sort of thing to sink in. And much of it is tied to party support, whereas what's important is how the fence-sitters see the debates. That's the lens through which I've been evaluating them.

  5. Oh, and just a suggestion for the graphics. Have you thought about putting the Bloc between the Conservatives and Liberals on the percentage projections? I think it would make the Quebec pie look more like the others, and easier to compare the strength of the parties between provinces. The Greens would end up as a sliver at the top in Quebec too, the NDP would be top right, the Liberals bottom right, and the Conservatives left.

    ...hey, maybe you should flip the pie chart over, too, so the Conservatives are on the right! :D

  6. Eric, I was wondering what you are going to do in the last week of the Campaign. The model is actually a trailing indicator of opinion when there are a large number of polls, and the relatively rapid movement of opinion during a general election. If you are giving weight to the polls from early in the writ period, and there is a consistent trend over multiple polls, then you will likely end up with a significantly incorrect prediction on E-Day.
    I would say that you will need to start adding strongly worded caveats to your post. What do you think? Would it make sense to discard older polls more quickly?

  7. William,

    That is a good idea.


    No, not based on tests for the 2006 and 2008 election. Polls should be coming more and more now, so older ones will drop off more quickly.

  8. Éric

    I agree that it can take days for the results to become clear, but the Ipsos numbers are intersting in that they are collected in the hours immediately before the debate and in the minutes immedaitely following the debate. They include the responses of uncommitted voters on who did best, so they tell an interesting story.

    I think the results are reported poorly. They are presented as a strtaight horse race, without standardization for uncommitted responses. Top line numbers don't mean much. But the response of the Don't know/Refuse category are potentially illuminating.

  9. Eric,

    Fantastic stuff. I tune into this site regularly from Australia --- find it to be a beacon of objectivity that is refreshing ....

    A question - why do you think Nanos has the Greens at such low numbers -- the other pollsters seem to have them pegged from 7-11 -- Nanos generally well below 5. Are the others using prompts?

    Just wondering

  10. Sure, and I encourage people to look at them and form their own interpretations. If I had more time I would probably look at them as well, but I over-estimated the amount of time I would have during the campaign to look at side issues and polls.

  11. Dookie,

    Thanks. Nanos doesn't prompt for any of the parties, which is the primary reason why the Greens get such poor results in their polls.

    Those who do prompt tend to over-estimate Green support, so the reality is likely somewhere in between.

  12. Hi Éric,

    I am questioning how accurate the riding information is. I can only assess my own riding, Pontiac, but the NDP is being given 19% support which has a) never happened before, and b) is unlikely given they named a candidate at the very last minute. Is there a breakdown of how these numbers are attribute that you could point me to?

    Many thanks for the great work


  13. Under the LE DEVOIR image is a list of links explaining how the site works. This one would be most relevant for your question:

  14. Eric, the site is fantastic, thanks very much for all your work.

    A question or two for you: We're always hearing when the punditocracy debates various polls and the media coverage thereof, that because phone polling is done via landlines exclusively and not cell phones, it tends to underrepresent younger voters. Do you give credence to this view? Do you factor in a correction for it into your model, or does the lower likelihood of younger people to vote at all make it a wash?

  15. Stephen,


    To answer your questions, more and more pollsters are including cell phones in their polls. I know for sure that Nanos and EKOS do, and I think most of the others do as well. It's still harder to reach them, but I think your speculation that it's a wash based on turnout is likely an accurate one.

    I don't make any corrections based on how the polls are conducted, only based on the pollsters' track record. The projection is also slightly adjusted based on how the pollsters have been off in previous elections.

  16. I think you are overestimating the impact of the French Language debate and overestimating Ignatieff's performance.

    Yes he did better than Tues. night but nothing I've read said he presented himself as the clear Federalist choice in Quebec.

    Also interestingly, if the numbers get close to the 2008 election, Harper's argument that the election was for nothing and do you want this to happen again gets stronger.

    More negative ads will be released in the bottom half, like them or not. I'm not convinced that in the long run the debate does anything to the numbers.

    If you go by the last election post debate polls, Dion came within 3 points of Harper, and we all know how that turned out.

    btw I think the Liberals poll higher on hockey nights in Canada. Just an unscientific observation.

  17. What about the Nanos poll that came out showing the conservatives at 49% support in Atlantic Canada?

  18. "The Liberals trail the New Democrats by 2.5 points or less in Trinity - Spadina and Sudbury"

    What a coincidence; where I live and where I used to live. Obviously both appearing randomly in a sentence is a sign of...something.

    I think NANOS today suggested that the NDP had made good on their hopes of gains from the Bloc, though I'd be curious to see if that helped the NDP or the CPC more; Or perhaps the NDP competing better against the LPC in Montreal.

  19. Anonymous 10:56,

    Yes, it is included in this morning's projection.

  20. Hi Eric

    I am going to mention EKOS so this likely won’t pass your moderation

    The EKOS analysis of the poll released April 13 ends with:

    “All in all, the race is reaching a very interesting and dynamic midpoint which could still yield some major surprises on May 2nd”

    OK does EKOS have faith in this poll that has the CPC leading the Liberals by only 5 points? Your seat projection model would use this poll to show a very weak CPC Minority 110 - 100 or so?

    As EKOS has published this poll why would they think that result would be a surprise?


    I saw that you gave the COMPAS poll sample size 2,251 a weight of .21 and the Ekos poll sample size 1,108 a weight of .63. Basically the opinions from the EKOS sample are worth 12 times more than the opinions in the COMPAS sample.

    AND you do not tolerate questioning pollsters integrity?

    I saw that you were rationalizing your weightings yesterday with EKOS being the 5th most accurate pollster. What are the top 5? Does the #1 most accurate pollster have a weighting factor 5X that of EKOS?

    If there is something obviously wrong with the COMPAS methodology please point it out and just ignore it.

  21. Today's Nanos is very interesting - the Bloc and the Greens are both at their lowest since the daily tracking started (7.5 and 3.1). Oddly enough none of the other parties registered their highest since daily tracking started, though NDP's is best since the first day of daily tracking.

  22. Seems to me that the main take away from the French debate is that it was a total fiasco for Harper. His numbers plunged and he was rated a distant fourth. It will be interesting to see if Tory numbers in Quebec take a hit after that miserable performance.

  23. I am surprised to see London West still listed in the Conservative column. This is a traditionaly Liberal riding that was only barely won by Ed Holder in the last election, again, by vote splitting between Lib/NDP. With the current rise in Liberal support across Ontario, I would be shocked if Holder can win again in London West. To make matters worse, his campaign has been dismal and scandal ridden and he is facing a well-liked and popular local candidate for the Liberals, Doug Ferguson.

  24. BCVoR,

    Please don't question my moderating policies. I have to avoid being sued, and people saying this pollster or that pollster is a fraud is the kind of stuff I can't allow on my site. I'm responsible for the comments on this site, something people need to remember.

    As you didn't say anything like that this time, I'm happy to allow it.

    COMPAS had a lower rating for several reasons. Firstly, it was taken over a longer period of time. As I weight polls by the mid-point of their field date, that reduced COMPAS's weight compared to EKOS, Nanos, and Innovative.

    For example, COMPAS was weighted as being three days older than Nanos's poll.

    Secondly, COMPAS's track record came into play. They were relatively close in the 2004 federal election but way off in the 2005 Alberta election. That lowered their weighting.

    Your calculation for the weighting of the polls is not correct.

    As to the top pollsters, I'm not going to release that list and I said EKOS was in the Top 5, not the 5th best. And no, the top pollster does not have 5x the weighting of the 5th ranked pollster.

    The top ranked pollster has a rating of 1. For each pollster, the difference in their average error from that of the best pollster determines their weighting.

    For example, if the best pollster has an average error of 1.5, and another has an average error of 3.0, that second pollster will be rated a 0.5 (1.5/3.0).

    All pollsters are weighted according to this criteria, so they are all treated equally.

    All of this has been explained elsewhere on the site, which you should know.

  25. Excuse me Eric but you may need to go back and scrub some past comments then.

    Over the last couple threads you've approved comments by people suggesting Sun media cooked up this Compass poll (ie. that its a fraud.)

  26. Please point them out. You don't know how many COMPAS-related comments I had to block.

    I can allow an expression of opinion if it is worded in a certain way.

  27. And if anyone is wondering, weighing pollsters by their track record is something Nate Silver has done on FiveThirtyEight as well.

    No subjective judgment is inserted in those rankings - they are completely objective and fact-based.

  28. Hi Eric - what are you doing about the fact that Nanos does not break out Alberta from the other Prairie provinces? It looks to me like you're just not including them at all. Is that true?

    If so, do you see that having some effect on the results that you project there?

  29. Ashley,

    I don't include Nanos's Prairie numbers, but I do include them from the other pollsters, virtually all of whom separate Alberta from the two Prairie provinces.

    It may cause my Green numbers in those three provinces to be slightly high, but I don't expect it to have any major effect. There aren't a lot of seats that are likely to change over.

  30. Hey Eric, just want to say how you have done a great job. I visit your website a couple times a day and it is by far the best website on the internet for Federal Election 2011. I especially appreciate that you take the time to take questions - a lot of the questions you've answered in this thread are questions I was also thinking of. Overall, great job, keep it up, it is much appreciated. I can only imagine how time consuming it must be to update multiple times a day - I hope you don't have a day job!

    I am still not sure where my vote will go but I would like it to be a Con minority ideally. Good to see that your projections has that type of result.

  31. Predicting seats from polls is a tough business. Nate Silver and 538 has way more local and state polls to assist him, than does Eric. If we had riding polls or even properly done provincial polls then we could argue with Eric's individual riding choices.

    All that I have seen from those who argue with his riding by riding picks is anecdotal evidence which while it may have some truth to it isn't in any way scientific.

    I'd like to see a majority government preferably by the CPC. I'd take the Liberals though. Right now it isn't there. There's no point in blaming Eric because the polls don't show what you want. I'd love to believe the Compass poll. Something tells me it is an outlier. Maybe I'm wrong. Again Eric simply pointed out that it's numbers needed to be confirmed by other polling.

    The constant squabbling on here reminds me too much of Parliament. Accept the results and comment on them rather than attacking the person who gives you the results.

  32. Interested in NQW14 April, 2011 15:23

    Do you poll in the individual ridings? I'm working my butt for the incumbent, but I'm finding a lot of the support we used to have locked down has eroded to the Libs. I would be very happy if you told me the prediction you have for NQW is accurate and up to date.

  33. Interested,

    I don't conduct any polling, I aggregate the polls of other firms.

    The projection I have for each riding is based on province-wide trends and individual factors related to each riding. When riding polls exist, I include them, but there have been no riding polls conducted outside of Quebec that have been publicly released.

    I have no specific data for Northumberland Quinte West.


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.