Friday, March 23, 2012

Two more federal polls, Conservatives lead by three or eight

After the headline-grabbing poll from Environics yesterday, Angus-Reid and Harris-Decima submitted their entries in order to further muddy the waters. Both show a Conservative lead, but the size of it varies considerably. They both show the New Democrats at just under 30% support, which seems to be the consensus across every poll, but their position in Quebec remains uncertain.
Let's start with the poll from Angus-Reid, which is the most recent. The firm was last in the field 20-21 January and since then the Conservatives have slipped two points to 37% support.

The New Democrats trail eight points behind with 29%, up one point since the January survey.

The Liberals are down one to 21% while the Bloc Québécois is up three points to 8%. The Greens have dropped one point to 4%.

The Conservatives lead in every region except Quebec: 40% in Ontario (-2), 43% in British Columbia (+9), 56% in Alberta (-9), 36% in Atlantic Canada (+6), and 51% in the Prairies (-21).

The New Democrats lead in Quebec with 33%, unchanged from the January poll. They are running second in British Columbia (33%, -7), Alberta (25%, +8), Atlantic Canada (34%, -4), and the Prairies (32%, +12).

The Liberals placed second in Ontario with 30%, an increase of one point, while the Bloc Québécois is second in Quebec with 31%, a gain of eight points.

It is interesting to note that, despite the Conservatives leading and only slipping slightly, Stephen Harper's personal numbers have taken a tumble: 37% of respondents said their opinion of him has worsened. That is more than double the 18% who said the same of Bob Rae.

The Angus-Reid survey also included some NDP leadership numbers, in terms of which candidate would make Canadians more likely to vote for the NDP. No surprises in the results: Thomas Mulcair got 12% while Brian Topp and Peggy Nash placed second with 4%. The rest of the candidates were slightly behind, but we're talking margin of error differences.

Regionally, Atlantic Canadians appear giddy to vote for the NDP no matter which candidate wins, as everyone but Mulcair had their best result in that region. Mulcair had his best result in Quebec (34%), while Topp had his second best in the Prairies (8%), Nash in Alberta (5%), Paul Dewar in the Prairies (5%), and Nathan Cullen in British Columbia (7%).
Now to the Harris-Decima poll, which was in the field the day before Angus-Reid, but also the 11 days before that. Harris-Decima was last in the field 12-22 January, so at about the same time as Angus-Reid's previous poll.

Since Harris-Decima's last survey, the Conservatives have dropped one point to 31%. They hold a lead of three points over the NDP, down one point to 28%. The Liberals are down one point to 24%, while the Bloc Québécois is up three to 8%. The Greens are unchanged at 7%.

The Conservatives lead in Ontario with 33% (-2), Alberta with 58% (-3), and the Prairies with 47% (+12). They are running second in British Columbia (33%, +3).

The New Democrats lead in British Columbia with 35% (-7) and Atlantic Canada with 34% (+7), and trail in second in Quebec (26%, -6) and the Prairies (31%, -4).

The Liberals are second in Ontario with 30% (-4), Alberta with 14% (-3), and Atlantic Canada with 33% (-1). The Bloc Québécois leads in Quebec with 34%, a gain of 12 points.

These are not exactly consistent results. Harris-Decima is more in line with Environics's poll, while Angus-Reid is more in line with the other polls that have been out the last two weeks. What to make of it?

The field days are quite different, in addition to the methodologies employed. Both show general stability and the truth likely lies somewhere in between their results, but at this stage opinion seems to be still somewhat foggy with the next election so far away and the NDP leadership campaign waiting to come to a close. It seems safe to conclude, however, that the NDP and Conservatives are probably not tied after all.

The differences make us search for consistencies. Both Angus-Reid and Harris-Decima were in the field around the same time in their last two surveys, giving us the ability to look at some trends. Are there any?

Nationally, we see the Conservatives and the Liberals both taking a step backwards. This seems consistent with some other polls, where we have seen the Liberals moving away from the mid-20s and the Conservatives from the high-30s.

Both polls show the Conservatives losing ground in Ontario and Alberta, while gaining in British Columbia. The New Democrats have lost since January in both polls in British Columbia, but have gained in Alberta. The Liberals are down in both polls in Alberta and Atlantic Canada, while the Bloc Québécois is up significantly in Quebec. That seems to be the main point of agreement across every poll.

More generally, they both show relatively close races in British Columbia (between the Tories and the NDP) and Atlantic Canada (between all three parties). They both show the NDP competitive, but still well behind, the Tories in the Prairies. And they all show the NDP running third in Ontario, and in a neck-and-neck race with the Bloc in Quebec. That last bit of agreement has the most important implications.

The seat projection for these two polls shows why. One would expect the Harris-Decima poll, with the narrow margin between the NDP and the Tories, to have the best result for the Official Opposition. But no - trailing the Bloc by eight points in Quebec means the NDP's ranks in the province are decimated. Holding steady at the national level is all well and good, but if it is the result of small gains in the rest of the country making up for losses in Quebec, the NDP just doesn't win the seats required to make good the drop in the province.

Angus-Reid's results would give the Conservatives 142 seats, the New Democrats 83, the Liberals 58, and the Bloc Québécois 25. The Greens were too weak in British Columbia to re-elect Elizabeth May.

Harris-Decima's results would give the Conservatives 127 seats, the New Democrats 67, the Liberals 67, the Bloc Québécois 46, and the Greens one.

Quite a difference in results. Angus-Reid keeps the NDP in the Official Opposition role but unable to combine with the Liberals to out-number the Tories. Harris-Decima puts the NDP's status in question, while giving the Liberals and New Democrats enough seats to out-vote the Conservatives. But in both cases, the support of the Bloc Québécois would be needed for any working coalition.

The regional breakdown is as follows, with Angus-Reid first and Harris-Decima second:

Conservatives win 21/17 seats in British Columbia, 26/27 in Alberta, 20/19 in the Prairies, 58/46 in Ontario, 8/4 in Quebec, 8/13 in Atlantic Canada, and one in the north. Their 338-seat total would be 159/142.

The New Democrats win 11/13 seats in British Columbia, 2/1 in Alberta, 7/6 in the Prairies, 22/26 in Ontario, 32/12 in Quebec, eight in Atlantic Canada and one in the north. On the 338-seat map, the totals would be 89/73.

The Liberals win 4/5 seats in British Columbia, 1/3 in the Prairies, 26/34 in Ontario, 10/13 in Quebec, 16/11 in Atlantic Canada, and one in the north. In the expanded House, they win 64/74.

If we take the best and worst regional results for the parties in these polls, we get an idea of a high and low range: 121 to 148 seats for the Conservatives, 61 to 89 seats for the New Democrats, and 53 to 72 seats for the Liberals.

In this context, it is difficult to see the Harris-Decima poll as a positive one for the New Democrats. They would probably be better off with the results from Angus-Reid, despite the wide gap between themselves and the governing Conservatives.

The polls certainly do not provide a lot of clarity as to what is going on. This is why I intend to start tracking the polls and maintaining an on-going vote aggregation (a more detailed version of my monthly averages), for the sole reason of making some sense of the numbers. I hope to have this up and running soon.

I had promised an update to the NDP endorsement rankings, but the only new endorsements to emerge since Wednesday were the six Nova Scotia MLAs plumping for Thomas Mulcair. I'll instead provide an updated tally tomorrow morning, when I begin my live-blogging (though here from home) of the NDP leadership convention, starting at around 9:30 AM. I intend to have plenty of charts and numbers to give you throughout the day as we track what is going on in Toronto.

19 comments:

  1. You could probably form a coalition with the Harris Decima numbers... the Liberals + NDP would still combine for more than the Conservatives. You'd just need the Bloc to abstain.

    ReplyDelete
  2. Regardless of how you view these results the clearest view seems to be a slow drop in CPC support.

    Eric can tell us but I've got a hunch this is pretty normal for a majority Govt of any type ?

    ReplyDelete
  3. Surely your seat predictions for each poll for the Conservatives in Atlantic Canada, and therefore your regional max/mins, are backwards. It should be 13/8, not 8/13.

    ReplyDelete
  4. Bruce Anderson said it all on At Issue last night, something to the effect of:

    For a pollster this is embarrasing. The numbers are all over the place. Somebody has to be wrong.

    ReplyDelete
  5. Recently I've been wondering about the BQ resurgence, and to a lesser extent that of the PQ. Could this fairly sudden change be caused by a reaction to the so-called Robocalls (or election fraud -- allegedly)? Could Quebecers just be sick of the whole mess and want to be rid of any attachments to Canadian politics? Just wondering.

    ReplyDelete
    Replies
    1. no... Federally, Quebeckers are waiting to see who becomes leader of the NDP - the holding pattern is to provisionally put support back with trad. BQ. Provincially, it's because the Liberal government is hated, and the CAQ has been outed as right wing. If Mulcair wins, support will go back to the NDP.

      Delete
  6. This earlier Léger pole, seams in line with Harris-Decima. Have you considered this?

    11 mars 2012
    Le Bloc québécois déloge le NPD
    http://tvanouvelles.ca/lcn/infos/national/archives/2012/03/20120312-044747.html


    Gut feeling, but Angus-Reid does not seam to fit with anything that I have seen in the last few weeks or the definate trend with the Parti Québécois that you show at the provincial level.

    ReplyDelete
    Replies
    1. I wrote about that poll earlier, it was part of a pan-Canadian Léger poll.

      Delete
    2. Then how do you explain the Angus-Reid results out of nowhere? completely out of tendency!

      Delete
    3. No, it is within what we'd expect - the Bloc is at 31% in both polls and the NDP is within the MOE. The two parties seem to be around 30%, some polls show one part of the other a few points up, a few points down. That is the trend right now.

      Delete
    4. This HD?
      31-28-24?

      Tory, NDP, Liberal
      37-29-21 AR - Mar 18-19
      30-30-20 Environics - Mar 6-18
      37-28-20 Abacus - Mar 9-13
      35-30-20 Ekos - Mar 6-11
      37-29-23 IR - Mar 6-8
      34-26-24 Leger - Feb 28-Mar6
      36-25-30 Nanos - Feb 25-29
      36-28-26 Forum - Feb 6

      Not just sure why you think the HD is closer than the AR.

      1 of these things (well 2) is not like the other. And if you are going to depend on HD as your weathervane,... come 2015 you are going to be very disappointed again.


      What I see based on the last month (or 2) is Tories in the high 30's, NDP holding just below 30, and the liberals holding just above 20. ....In other words we are still very close to a majority, especially if (proper) adjustments weren't made to the likely voter models.

      Delete
  7. Eric

    Your seat prediction model seems to be off.

    The AR poll is almost exactly the same as the last AR poll that was done before the last election April 29,2011

    cpc 37 Liberal 19 NDP 33.

    You have the CPC with the same popular support (according to AR) and losing 23 seats with exactly the same support they had 2 days before the last election.


    Are you making an attempt to incorporate the 30 new seats into your model???



    Is your purpose to have a mathematical/statistical exercise to translate polls to seats or to try to correctly predict seats based on polls?

    ReplyDelete
    Replies
    1. BCVOR - You're ignoring the differences in the regional numbers. The last Angus Reid poll had the race as 41/27/26 between the Tories, NDP and Liberals in Ontario. This Angus Reid poll has it as 40/24/30. That's probably where the seats are flipping from Conservative to Liberal.

      Delete
    2. Ryan is right, seat projections are made at the regional/provincial level, not the national level. That is why the NDP does so much worse in the HD poll despite the national margin being smaller.

      Delete
    3. Ryan --- Excellent fact checking.... I sure won't get in a heated discussion with you without having my facts straight.

      So a drop of 1 % pt in CPC and an increase of 4% for the Liberals 3 at the expense of the NDP will cost the CPC 24 seats in Ontario? And the NDP will some how hold onto 18 of their 22 in Ontario? In Eric;'s seat allocation lose 25 overall and 21 to the BLOc in que so only a max of 4 in Ontario.

      The real reason is that Eric makes no attempt to tie the pollsters previous record to the projections. All of the sudden when AR says 37% it means that the CPC will get 37% rather than the historical more accurate 39.6%.

      He is just doing a mathematical excercise assuming that the data provided to him by the pollsters is somehow spot on dispite it not being so for at least the last 2 election cycles.


      I believe the same in when it goes against the CPC... If COMPASS comes up with a poll with cpc at 46% it likely will give the same result as last election. If the have the CPC at 40% it almost surely means the CPC do not get a majority.

      KISS

      Delete
    4. When I make a projection for a poll, OF COURSE I USE THE ACTUAL NUMBERS FROM THE POLL.

      When I make projections based on aggregation of all polls, they are weighted by the track record of the pollster and adjustments are made according to where the party sits in the legislature. This, for instance, is why the Alberta NDP is projected to get 12.7% in my Alberta projection despite polling at 13% to 14%.

      But if I was doing a projection for an individual poll, I would obviously use 13% or 14%, whatever that poll indicated.

      It is absolutely ridiculous and nonsensical to suggest that I should be tweaking polls to and fro based on one or two data points, particularly when no poll was conducted on May 2, the actual day of the vote, and when the results were, in many cases, within the margin of error. Angus-Reid could have done its last poll of the campaign five times and gotten five slightly different results - it even happened in the Ontario campaign!

      So enough about this.

      Delete
    5. BCVR is just desperately trying to find a way to spin this in Harper's favor. He's having trouble with the fact he is slipping, which is made very clear when you see how his personal numbers are plummeting.

      Delete
  8. An aggregation is a good idea. A running 2 or 3 month mean would be good too.

    Also useful would be a plot for individual pollsters so the consistent bias could be easily spotted and reported.

    Lastly, I find it surprising that there is no analysis here of polling methods and "push-polling", asking a sequence of questions that elicit a biased response, either deliberately or not. For example, asking about the performance of the PM before a question on who you would vote for could elicit a different response than asking the question of voting preference first.

    Obviously polling companies are doing things differently. It would be interesting to see what methods are responsible.

    ReplyDelete
  9. Tory, NDP, Liberal
    31-28-24 HD - Mar 18-19
    37-29-21 AR - Mar 18-19
    30-30-20 Environics - Mar 6-18
    37-28-20 Abacus - Mar 9-13
    35-30-20 Ekos - Mar 6-11
    37-29-23 IR - Mar 6-8
    34-26-24 Leger - Feb 28-Mar6
    36-25-30 Nanos - Feb 25-29
    36-28-26 Forum - Feb 6

    .... aside from a rogue environics poll, and the usual broken ougee board that HD uses.... What is so different between the other polling companies??

    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.