Sunday, March 23, 2014

Sunday Prediction - Canada

Given that there have been 0 polls since the Debate, I decided this Sunday I'll do a prediction for Canada.

No real analysis. I simply used polling averages and trendlines.

The polls I used were as follows:

50% Lib
24% CPC
22% NDP
3% Grn

36% Lib
31% NDP
15% BQ*
14% CPC
3% Grn

41% Lib
30% CPC
23% NDP
5% Grn

42% CPC
30% Lib
22% NDP
5% Grn

50% CPC
26% Lib
15% NDP
6% Grn
2% Oth**

British Columbia
29% - Lib
29% - CPC
29% - NDP
12% Grn

*This presumes the Bloc is all but dead. Polls showing the party at 25% will be wrong - I predict - not because a raw 10% will swing to another party, but rather, they will stay home. Turnout in Quebec will be much lower by comparison.

**This presumes that, if Wildrose does not run Federally, that other right-wing alternative candidates (Christian, Libertarian, etc) will get these votes.

144 - Liberal
112 - Conservative
80 - New Democratic
2 - Green


  1. Note that Outremont is not an error; but I also did not heavily math-test this.

  2. This comment has been removed by the author.

    1. I question your "poll averages"; the highest level of support the Liberals have polled in Ontario since Trudeau became leader is 38%. An average of 6 polls for February place the Liberals at on 37.4% and the Tories at 33.7%. In Quebec the Liberals are polling 4-8 points below the 36% you have for them.

      Only two polls since the 2011 election have placed the Liberals at 40% or better; Forum research Oct. 23, 2013 40%, May 22, 2013, 44%. Most polls place the Grits in the mid-30's.

      I do not see a list of polls or polling companies used to compile these "averages". Without some corroboration of the polls used the numbers appear to be fictional.

  3. I find your "prediction" very odd for a number of reasons.

    Firstly if the Tories received 20% in the Atlantic and the Liberals at 50% Conservatives would be lucky to win 1 or 2 seats. Teddy, you have given them 6. In the 2004 election when the Tories averaged 30% in the four Atlantic provinces they won 7 seats; in 2004 the Liberals averaged 46.2% in the Atlantic and a total of 22 seats, you give the Liberals 23 seats on a 50% "average" and the Tories 6 seats on a 20% "average". This does not make sense. My own algorithm would give the NDP 3 seats and the Liberals 29 when I punch in your numbers. My algorithm is not perfect I admit but, from a logical perspective it is difficult to see how 20% popular vote for the Tories leaves them with 1 seat less than 2004 when they scored 50% better.

    British Columbia is also very odd to the point of skepticism. You give the Greens 12% and two seats; Elizabeth May's and Victoria. While it is not outside the realm of possibilities for the Greens' to capture Victoria an extremely and unexpectedly large 20% swing from NDP to Green would need to occur. This is equivalent to the NDP losing 40% of its vote to the Greens in Victoria-Not impossible I suppose but, very unlikely. However, on the 12% you give the Greens in BC it is essentially impossible to have the Greens winning Victoria especially when you give the NDP 88% of their 2011 vote in B.C. Even if the entire 5.2% increase in Green's B.C. popular vote happened in Victoria it would be essentially a toss-up whether Victoria remained NDP or went Green. We know however, that increases in popular vote are unlikely to be limited to one particular riding or geographic area. It is illogical to assume a 66% increase in the Green vote as you predict, would all be concentrated in Victoria-nearly impossible and certainly extremely improbable.

    Finally, you write that you use "polling averages and trend lines" what is it? Presumably you can't use both for a particular region, you must use one or the other. In Quebec for example you have the BQ at 15% not impossible but, in the last year the BQ has not polled below 4% nationally which would equate to slightly over 15% within Quebec. In February the BQ averaged 5.9% or 22.4% in Quebec.

    You write: "No real analysis. I simply used polling averages and trendlines (sic)".

    Sadly, I can not find any polls that would create the numbers above. The numbers for the Liberals are below their average in Atlantic Canada and are completely opposite to the trend line we see in the Atlantic provinces. Quebec appears to be more assumptions since the polling averages do not fit the "averages" you have assigned to the parties. That pattern holds throughout Canada with the exception of B.C. which do appear to be more or less consistent with polling averages.

    Unfortunately, I must conclude without further proof such as a list of polls used to construct "averages" and "trend lines" that you have simply used arbitrary numbers then through educated guess work determined what ridings would be won by a particular party. This is disingenuous at best since you portray your numbers as being based in fact! If the numbers are based in fact you should have no problem proving my assertion incorrect!

    1. The ElectoMatic and it's ratio system were used for the predictions.

      The Ratio system works in the following way.
      Lets say Party A took 10% in a riding last time, and 30% in another riding, and, 20% provincewide.
      Now, the party is polling at 40% provincewide.
      40% is double 20%, so we double the riding numbers. 10% becomes 20% and 30% becomes 60%.

      Now of course it is not this simple, you need to re-add all the ridings to you end up with some over 100% and some under 100%, but that's the simple principle that has lead me to being the closest to the mark in the 2011 Federal election for anyone doing individual riding by riding projections.

      You noted I used the word "Trendlines", this means that I adjust the numbers based on trends I see.
      One of the reasons I've been more successful at making mathematical projections than many others, is that I do not use simple raw math, I use my "gut" to fill in the gaps.

      My gut tells me that there is no way the Liberals will actually take more than 50% in the Atlantic, despite the polls suggesting they will. My gut tells me the Liberals will go up in Ontario. My gut tells me the Bloc will go down in Quebec.

      Remember as well this is a prediction and not a projection, it's held to a lesser standard by myself. If it was the latter, the Liberals would not be at "41%" in Ontario, but rather, 40.98% or 41.22% or something. There would also have been much more weighting of polls based on date and company and sample size and what was going on when they were taken, etc.

      The ElectoMatic is free. You can get it here
      This is the "Lite" version. That means, among other things, that the final calculation - the one that makes sure each riding adds to 100%, - is not done. I'm still working on a more robust version. The end result in terms of who wins each seat, however, is the same regardless of version.

      The file comes pre-loaded with the last calculation I ran - in this case, the ones used to make this map - but feel free to adjust any of the numbers as you see fit.

    2. Note that (upon opening the file) I made some adjustments to some ridings (like Outremont) and thus, ridings like that will not match the maps.