Friday, May 8, 2009

New Projection: Harris-Decima

As I mentioned on Tuesday, Harris-Decima released a new poll. I have the details, and so I have updated the projection. The poll was taken between April 23 and May 3, and involved 2,026 interviews. For the results of the poll, see the preceding post or scroll to the bottom of the page.

First up, here is what the electoral results would be of this poll, with the difference from the most recent previous poll in brackets:

Liberals - 126 seats (-11)
Conservatives - 105 seats (-6)
Bloc Quebecois - 52 seats (+6)
New Democrats - 24 seats (+10)
Greens - 1 seat (+1)

This poll would have given the Conservatives 60 seats west of Ontario, 37 seats in Ontario, none in Quebec, and seven in Atlantic Canada. The Liberals would have 24 seats in the West, 57 in Ontario, 23 in Quebec, and 20 in Atlantic Canada. The NDP would have had eight seats in the West, 12 in Ontario, none in Quebec, and four in Atlantic Canada. The Greens would win their seat in Atlantic Canada.

The short-term projection taken from the last five polls has changed as well. First, the popular vote with difference from last update in brackets:

Liberals - 34.2% (-0.5)
Conservatives - 31.4% (-0.2)
New Democrats - 14.8% (-0.1)
Bloc Quebecois - 9.2% (-0.1)
Greens - 8.6% (+0.6)

And now the seats:

Liberals - 128 (+1)
Conservatives - 114 (unchanged)
Bloc Quebecois - 48 (-1)
New Democrats - 18 (unchanged)

As for the long-term projection, there have been no seat changes. The national popular vote has changed:

Greens +0.3
Liberals +0.1
NDP and BQ unchanged
Conservatives -0.5

There have been some significant changes in the regions:


Conservatives -0.6
Liberals -0.1
New Democrats unchanged
Greens +0.4


Conservatives -0.9
Liberals +0.7
New Democrats -0.2
Greens +0.5


Conservatives -0.6
Liberals -0.1
New Democrats +0.2
Greens +0.8


Conservatives -0.1
Liberals +0.1
New Democrats unchanged
Greens +0.5


Conservatives -0.5
Liberals +0.4
New Democrats -0.2
Bloc Quebecois unchanged
Greens +0.3


Conservatives -0.3
Liberals unchanged
New Democrats -0.2
Greens +0.3

So, because the Liberals have been unable to bury the Conservatives in Ontario as of yet, they are still ahead in the long-term projection. But there is a trend and if the Liberals surpass the Conservatives nationally in the long-term projection, which they look to do soon, the projection will likely swing to the other side dramatically.


  1. How do you determine the weights attached to each poll and election result?

  2. Three factors:

    1) Age of the poll. For simplicity, I classify polls by month. Polls from this month and last month are weighted relatively high, afterwards they reduce in weight by 33% per month.

    2) Reliability of the polling firm. This is determined by how accurately the polling firm guessed the electoral result in 2008. I used the closest as a starting point, and reduced each subsequent polling firm by how far they were from the closest.

    3) Size of the poll. Here again for simplicity, a poll including 1000 people will not add or subtract from the weight of a poll. A poll with 2000 people will have twice as much weight, a poll with 500 will have half as much.

  3. As for the electoral result, I started with simple weights, with 2008 being 100%, 2006 being 50%, and 2004 being 25%. Since then, I've been reducing each of the weights each month depending on how many new polls have emerged.

  4. I'm very curious as to how the GPC will win a seat in Atlantic Canada. It is very pleasing to see the prediction, but unfortunately I have a good idea about the organizational strength in Atlantic provinces. Outside of Central Nova, there are some modest organizations at best. central Nova itself is pretty much a writeoff, unless, unless Petre MacKay slips on several dozen banana peels consecutively. (Or is hit by a bus).
    For a Green seat, I'd lay my money on Ontario, and most likely Bruce Grey Owen Sound, provided the Liberals show up with a decent candidate to split the vote for us.

  5. Surprises happen. The Conservatives won 10 seats in Quebec in 2006 without an organisation.

    The calculation is simple. If May needed to increase her vote total by X to win and the Atlantic region is up by X then May is calculated to win.

    It's obviously more difficult to project for the Greens considering I don't make my projections riding-by-riding.


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.