Saturday, November 26, 2005

Need an excuse to ignore polls? Here's one.

(I started writing a piece early on this morning when the papers came out about today's competing polls. My attempts at a side by side comparison were thwarted, however, because I can not get Blogger to act sanely with tables. Here's what I had saved as a draft:)

There are plenty of reasons. Even when polls are accurate, what they are measuring is fleeting. The wording of the questions often shapes the results. The numbers themselves are worth little without the proper context and analysis.

I follow them anyways. I like to have a general sense of where opinion is, and how it has shifted.

But here's a doozy:

The National Post today has a poll (Ipsos-Reid) it claims as a "tight" race.
The Toronto Star today has a poll (EKOS Research) showing the Liberals with majority government levels of support.

Both polls were conducted over the same time period (November 22 - November 24).

Here is how they compare --

Polling data comparison
partyIpsos-ReidEKOS
LPC34% (-3%)38% (+5.7%)
CPC30% (n/c)29% (+1.5%)
NDP16% (n/c)16% (-4%)
BQ15% (+3%)10%
sample size1,000802
margin of error3.1%3.5%


Ultimately EKOS and Ipsos-Reid numbers are different, but not dramatically so. If you take the margin of errors into account, they aren't so far off. What is more interesting is how the newspapers spin the numbers, and how the polls seem to so nicely fit into the overall philosophy of the organizations who commissioned them.

You find what you are looking for, I suppose.

Anyways, I obsess over polls a bit. I reserve my right to do so. But I pretend to myself that I follow them because I am concerned about how people are reacting to what happened yesterday, rather than as prophetic insights into what might happen tomorrow. Yesterdays are important. We learn from yesterdays.

Hopefully.

Something about Ipsos-Reid's BC numbers seem off to me (Liberals down 18% and Conservatives up 14%?). I'm not sure I like the EKOS sample size, so much, either, and the Bloc seems quite unbelievably low at 10%.

But the results aren't as far apart from each other as the headlines suggest. Other than divergent trends of a few percentage points among the LPC and BQ, they are practically identical.

One thing both do seem to show is that the Liberal support that bled off to the NDP has gone back to the Liberals and the Conservatives are - again! - stagnant.

Hmm. You know what? Your time is valuable, and the speculation isn't worth it. My suggestion for today is: read Mallick's latest column. It's a good read.

0 Comments:

Post a Comment

<< Home