Thursday, May 15, 2014


I wanted to address this issue head on given the very strange nature of the problem we as predictors and projectionists face here in Ontario.

We've had 8 polls since the campaign begun on May 1st.
Of those, 4 have indicated a PC lead of between 5 and 11 points.
And 4 have indicated a Liberal lead of between 3 and 6 points.

Both Ipsos polls show a PC lead, while both Innovative polls show a Liberal lead. Forum has one of each. 

Within each group of 4 there is at least one IVR, Online, and Telephone poll. Each has large (1.5K) and smaller (under 900) sample sizes. It's thus very strange that we would see two groups that seem to disagree in such a manner.

Within the polls showing a Liberal lead, the average is 38% for the Liberals, 33% for the Tories, 22% for the NDP, and 6% for the Greens.

Within the polls showing a PC lead, the average is 31% for the Liberals, 39% for the Tories, 25% for the NDP, and 5% for the Greens.

So, what is going on. Why such the different numbers? I decided to dig in to this to try to figure it out.

I compared regions; there was nothing useful that came out of that. I compared past polling from the same company, but that usefulness of that was debatable. I compared education, immigration status, age, but none of it seemed to fit.

Then I compared Men and Women.

All 8 polls - or rather, those within the 8 that detail the breakdown - agree that Men favour the PC Party by some margin, usually narrowly.


Polls showing a PC lead all indicated that Women distributed their votes in a similar pattern, while all the polls showing a Liberal lead had Women voting Liberal by wide margins. Compare for example the two Forum polls, showing differing results:

You can see the split among male voters is at 41-33 in one poll and 41-34 in the other. However, among Women the split changes from 35-33 to 30-42

Why this is, I do not know. I simply noticed this very odd pattern; something I've never seen before. I will start to dig in to past polls for past elections and see if I can figure out what this means.


  1. humm

  2. I cant find any solid numbers to back it up; but it would seem this inconsistency is caused by there being a large number of women that are very iffy about voting PC. If a pollster does not press an undecided, they thus do not show, but if they do, they admit they are thinking PC.

  3. I always find that Ipsos polls always over-inflate both NDP and Conservative support. In the polls showing a PC lead, the NDP support is at least 3 to 4 points higher then the other polls. As well, the Conservative levels more then likely are not quite as high either. Forum, although considering their poorer track record, where pretty much right on with the Quebec election results.
    Your observation about women voters is interesting. I have a hard time believing that most women would vote PC when historically, women are more likely to vote Liberal or NDP. Just some random thoughts on my end.

  4. Interesting theory, Teddy. I think most people would attribute it to "likely voters" and the general lack of enthusiasm there is for all the parties, but you could be on to something here.

  5. I wonder too if this has something to do with it:

    I wish we knew what time of day these polls were taken. There may be a huge difference between stay at home wives that have husbands with stable jobs, and single mothers struggling to find a job in the first place.

  6. Ok where does the "poorer track record" for Forum come from? They did one election really badly (the by-election in Brandon-Souris). Otherwise they have pretty much been the best in most provincial election. They do tend to have more volatility but I don't see as a bad thing, it probably just shows their samples are truly random.