Thursday, July 31, 2008

About polls on Morgentaler and the Order of Canada

About polls on Morgentaler and the Order of Canada

In the wake of the criticism of the KLR VU Research poll showing that 56% of Canadians oppose Morgentaler receiving the Order of Canada, I thought I would return to the Ipsos-Reid poll that said 2/3 of Canadians support Morgentaler receiving the award.

I emailed Ipsos-Reid to ask for details about the poll. The first mistake I made was to assume it was a telephone RDD poll-- your typical telephone survey. It was not. It was an online panel poll. It surveyed 1023 Canadians from its pool of potential respondents.

There's bias right there.

In order to be included in this survey, one would have to have a computer and internet access, and then have signed up to be in Ipsos-Reid's pool of respondents. That makes the sample somewhat self-selected. And this poll would have been biased towards those who are comfortable reading and writing English.

Ipsos-Reid did try to make the poll representative of the population according to the Census Data. But you can conceive of all kinds of populations that were missed in the poll.

It's not a truly random sample.

I'm also concerned about some of the weighting that occured. There were 143 respondents aged 18-34. When weighting was factored in, they were counted as 285.

I'm no statistician, but isn't that a lot? We know that young people tend to support legal abortion more than older people.

The 55+ crowd was weighted down from 467 to 331.

There's a whole lot of weighting going on. As I understand it, that would introduce a lot of variability. It seems to me that if the sub-group were weighted so it increased by a small percentage, that would be acceptable. But to double the size? In a poll of 1000 people, that could influence the outcome.

The Angus Reid poll (sponsored by the Toronto Star) was also an online panel poll.

This is a poll that said that 45% of Canadians feel that abortion should be permitted in "all cases".

I don't believe that statement for a second. LifeCanada's yearly poll on attitudes about abortion shows that about 30% of Canadians believe that human life should only be protected at birth. There was an over-representation of those who support abortion-on-demand. That would definitively affect poll results.

It also had a "not sure" option. The "not sure"'s were fairly high for a poll.

The media and the left-wing bloggers accepted these polls without question, without looking at the methodology.

It is true that the KLR VU poll was not, technically, a random sample of individuals. However, it was a random sample of households. The average Canadian had a far greater chance of being contacted by KLR VU than of being included in the Ipsos-Reid or the Angus-Reid poll. Each sample is drawn from a panel-- a database of possible respondents. From what I've read, the number of potential respondents per panel only numbers in the tens of thousands. Ipsos-Reid couldn't even get a representative sample of 18-34 year-olds to respond to this survey-- not even close.

I think this should be borne in mind when comparing the results of these three polls.

For more social conservative news check out