Site Administrator Of:

Supporter Of:

Archives

My new favourite (Canadian) electoral projection site

I’m not sure how I missed this site before today, but thanks to Steve for mentioning it over at his blog and alerting me to it.

Basically, the site does popular vote percentage and seat projections for the Canadian political parties based on the most recent 5 polls it enters into their database,  but it also will do seat projections based on specific polls, such as the Ekos poll that was released today.  In that specific example, their site projects these results:

This poll would result in the following seat totals:

Liberals – 125
Conservatives – 106
Bloc Quebecois – 50
New Democrats – 26
Greens – 1

I like the site enough that I think I’ll be adding it to my sidebar.

Share

4 comments to My new favourite (Canadian) electoral projection site

  • I do not make individual riding projections, but make my seat projections based on regional numbers. Will my site tell you who will win in Central Nova? No, but it will tell you what the next Parliament will likely look like.

    I used this system during the Quebec provincial election and was the closest projector. I was the only one to project the PLQ winning fewer than 70 seats and the PQ more than 50.

  • Simple Massing Priest

    The methodological problems with most projections – including at least the “short-term projection” on this site – are so huge as to make the exercise utterly meaningless. The inherent assumption is that popular vote shifts are consistent, either across all the seats in the country or at least across all seats in the region. That, of course, is completely false.

    A Canadian general election is not one election, but rather 308 distinct elections, each with their own dynamics. These types of projection sites, however, insist on treating it as one electoral contest (or at best five or six regional contests), resulting in projections that completely ignore local dynamics. Using this fundamentally flawed methodology, there was simply no way, for example, that John Turner could possibly have won Vancouver Quadra in 1984. Yet somehow, no one told the voters of Vancouver Quadra.

  • Ian

    With FPTP I tend to not buy into these popular vote based projections and prefer a detailed riding-by-riding regional breakdown like DemocraticSpace (which only provides during elections – also it looks like they’ve unfortunately been hacked recently). None of the predictors (or even media) called a NDP seat in Alberta, but DS did a few times.

  • Thanks Scott, I appreciate it!

    I actually have two basic projections. The one at the top of the page is the long-term projection, which takes into account past electoral results and polls stretching back to December. It gives each poll a specific weight based on the reliability of the firm, the age of the poll, and the amount of people polled. So, this long term projection is what I would call the “official” projection of ThreeHundredEight.com (that address will bring you to the blog, by the way), and the more accurate one.

    The second projection is based solely on the last five polls, and does not take into account size or reliability. That is what I call the short-term projection, as it is more representative of what the situation is right now, rather than what it will be when all is said and done. The long-term projection does not fluctuate widely when a new poll is released, as it is a more (small c) conservative projection. The short-term projection can fluctuate more quickly.

    I also do seat projections for each poll as they are released, which is what you referenced in your post.

    Thanks again!

unique visitors since the change to this site domain on Nov 12, 2008.