Go to the Globe and Mail homepage

Jump to main navigationJump to main content


Entry archive:

Wayne Smith, who replaced Munir Sheikh as Canada's chief statistician during the census controversy, gives an interview in his Ottawa office on Feb. 11, 2011. (Dave Chan/Dave Chan for The Globe and Mail)
Wayne Smith, who replaced Munir Sheikh as Canada's chief statistician during the census controversy, gives an interview in his Ottawa office on Feb. 11, 2011. (Dave Chan/Dave Chan for The Globe and Mail)

Is census data usable? 'Our thinking has evolved,' chief statistician says Add to ...

That is their method for conducting a census.

Is that viable in Canada? Would Canadians even conceive of doing something like that? Those are issues that have to be addressed.

Back to the census we have now. This process [the NHS]will obviously be more labour intensive and that's where the $15-million comes in.

Maybe. Remember the $15-million is a contingency amount. If we get to the response rates that we need we may never spend that money.

What are the response rates you need then?

The targets are 98 per cent on the census and 50 per cent on the National Household Survey.

But the 50 per cent ... one national number doesn’t do it for us. We need the response to be fairly uniform across all geographic areas and fairly uniform across all population groups.

So we're not aiming necessarily for exactly 50 per cent and then we stop when we get there. If the response rate in B.C. is 25 per cent and it's 75 per cent in Quebec, we will need to continue.

So it's the notion of trying to get to a relatively uniform rate of at least 50 per cent. If the response and participation is very high it is possible we may go well beyond 50 per cent without ever spending the $15-million.

As can you can imagine, we got a crash course in census response rates [last summer]and methodology while combing through all those documents [released under access to information] One of the documents last year said we expect a 50 per cent [initial response rate]and said if we go back [to the doors] we could bump it to 65 to 73 per cent or something like that.

Oh these things were in the access to information [release to media]... When we do an access to information request, we put a lot of documents out, and there are things that are well founded and there are things that are not ...

This was a brief to the minister's office.

A brief to the minister's office.

It was a March 2010 e-mail to Mr. Clement's office, a senior StatsCan official advised that a self-administered voluntary survey would yield an initial response rate of only 50 per cent. She added that with follow up work and sufficient resources the agency could push the response rate up to 65 or 70 per cent.

I just want to be clear on that.

Our planning assumption is 50 per cent and our target is 50 per cent. Is it possible that we could get to 65, 75 per cent? Absolutely. That statement was early days in the thinking about what we felt we could say competently.

So if you have a response rate of 50 per cent versus 94 per cent before and you're asking largely the same questions – give or take the shift in the language questions – what is the erosion?

How much do you lose in terms of richness of data?

We should talk about that fairly thoroughly, actually. As I talked about a minute ago, the major difference between the National Household Survey and the 2006 census long form is essentially that the survey is voluntary and the sample is bigger.

The one thing we know with absolutely certainty is the response rate going to fall from making the survey voluntary.

We haven't done it before this way so we don’t know by how much. Fifty [per cent]is a very conservative number. We're likely to get there relatively easily I would think -- but you never know. It depends on the environment.

If the response falls to 50 per cent, there's one inevitable consequence of that. And that's exactly why the sample has been bumped to one in three from one in five. To compensate for that effect to the extent possible.

At a 50-per-cent response rate with a one in three sample, the sampling error would still be somewhat worse in the National Household Survey than it was in the 2006 census, the long form.

If the response is actually 60 per cent on the NHS it’s a wash. They are basically just as accurate, one as the other.

And if it goes to 70 pct actually the NHS sampling error is smaller than it was in 2006.

Report Typo/Error
Single page

Follow us on Twitter: @taviagrant, @stevenchase



Next story




Most popular videos »

More from The Globe and Mail

Most popular