Comments on: Tracking the tracking polls

By: Tracking the tracking polls: Nanos becomes less of an outlier

Tracking the tracking polls: Nanos becomes less of an outlier — Fri, 26 Sep 2008 01:02:38 +0000

[...] my previous post – all of two days ago – I noted that Nanos was consistently publishing estimates for Liberal [...]

By: Whiskey Echo (N. Van, not affiliated or decided... yet)

Whiskey Echo (N. Van, not affiliated or decided... yet) — Thu, 25 Sep 2008 05:39:30 +0000

It really is garbage, having this many “rolling” polls. The news media are trying to cover their butts and always be “scooping” each other, but really, have you ever seen so many polls that meant so little? I have some academic knowledge of how polls are actually done, and these polls are the loosest and most meaningless kind. Yet they dominate the newscasts and the media’s questions.

Here in BC we were subjected to some lame-arse poll of “Who is the most charismatic leader” – including Obama and McCain! What kind of cooked-up malarkey is this? Are we really going to start including another country in our polls, just because they’re having an election too? How about Japan, or South Africa, or Lebanon too? It’s not even the same kind of electoral system – Prime Minsters & parliaments are not elected like presidents. Idiots!

The media’s been apallingly bad this election, the government record and issues are hardly talked dealt with, while all we have gotten is polls, puffins, sweaters, and skinny dippers. I tell you, the fix is in.

By: Stephen Gordon

Stephen Gordon — Thu, 25 Sep 2008 00:59:28 +0000

Yes, it should. But since the polling firms don’t make public how they calculate their standard deviations, I couldn’t correct for that – although I suppose I could have made some sort of rough ad hoc adjustment.

But I do agree that as far as polling accuracy goes, a 6-point gap for the Greens is a much bigger puzzle than an 8-point gap for the Liberals.

By: Aaron Kreider

Aaron Kreider — Wed, 24 Sep 2008 23:41:36 +0000

I’ve always wondered – shouldn’t the standard deviation be a lot larger for a party polling at 30% than one that polls at 15%?

It often looks like they are using the 1/sqrt(n) formula for standard deviations, while (if I remember my stats right) that only works if you have two choices.

By: Stephen Gordon

Stephen Gordon — Wed, 24 Sep 2008 21:24:40 +0000

Strategic Counsel is limiting attention to a small number (45, I think) of swing ridings; they’re not publishing national numbers.

By: RossK

RossK — Wed, 24 Sep 2008 20:21:40 +0000

Stephen–

Very insightful – thanks.

Given Mr. Nanos’ track record, I am a big fan.

However, re: the methodology in this particular case….I don’t quite get it because, after all, is it not multiple choice in the polling booth…..

____
One other small question – why no Strat Coun numbers?
Too few polls over this period?

Thanks again.