Harper's Numbers Going South/Horserace Numbers Odd

EKOS released some leadership numbers today, that show a decided erosion for Harper. I'd also like to address a statistical curiousity I found with the poll horserace numbers.

First the leadership tallies:

Approval rating – Harper:
¤ 33% approve
¤ 52% disapprove
¤ 15% do not know/no response

• Approval rating – Ignatieff:
¤ 22% approve
¤ 48% disapprove
¤ 30% do not know/no response

These numbers are better for Ignatieff than first blush would suggest. First off, since the last time EKOS asked this question Ignatieff has rose 3% on approval, fell 2% on disapproval, for a 5% narrowing. What is particularly encouraging, this finding clearly demonstrates that Canadians don't hold firm views on Ignatieff, he is very much still a work in progress, which in an of itself denotes opportunity- the large "don't know" contingent speaks to this dynamic. Ignatieff clearly has work to do, but perceptions aren't cemented.

When you look at Harper's numbers, you see a large swing from the last finding. Disapproval up 10%, approval down 7%, which translates to a 17% widening. By any measure, that's a decidedly horrible trend. The numbers are even worse, when one considers the rabid approval Harper gets from Conservative partisans. On the other hand, one could argue Harper enjoys a more loyal base following, relative to Ignatieff. Still, Harper is a relatively known quantity, which makes these numbers that much more informative, cemented.

----

I was reading the pdf for the horserace numbers, that show a Liberal Olympic bounce ;). EKOS is kind enough to release the day to day findings:


I note that February 19 shows a one off Conservative tally, which really helps their overall score. I'm not quibbling about that, but it is a "which of these things doesn't look like the other" proposition. That aside, take a look at the Liberal numbers for each day, because this is where I started doing some math. The Liberal overall total of 30.3% seems a tad low when you review the day to day totals. Okay, here's what I did, and maybe someone could point out my error. I took each individual day total of votes (for example 569 for Feb 17) and divided the party percentages into that number, to get a voter tally. For instance on Feb 17, the Cons received 32.2% which gives them 183 of the 569 people sampled that day. I did this for both parties, for each day, then divided the grand total by the grand total of voters sampled. In this way, the margin of error for each day is fully incorporated and weighted. What I found confirmed my initial questioning, it shows the Conservatives at 33.2%(33.4% reported) and the Liberals at 30.9%(30.3% reported).

Why this is important, is because you see this "Olympic bounce" argument evaporates (not that it was valid in the first place). The Conservatives still rise 2% poll to poll, but the Liberals rise 1.9%- in other words a complete and utter wash, nothing to see here, the dynamic unchanged. You tell me, because I can't see how a party averages their lowest daily score, when they had days well above??