Evidence of meeting #61 for Industry, Science and Technology in the 40th Parliament, 3rd Session. (The original version is on Parliament’s site, as are the minutes.) The winning word was data.
A recording is available from Parliament.
Chief Statistician, Statistics Canada
I don't think we should burden the Canadian population more than is needed to get the data for the purposes for which it's intended, so I think we have to look at each case on its merits.
NDP
Brian Masse NDP Windsor West, ON
Yes. It's an interesting philosophy, though. If it made sense to do this, we'd do it for everything; if not, then obviously this is a weaker case, and the national survey won't stand the test that the census will.
Chief Statistician, Statistics Canada
Parliament addressed that issue in 1981 when it decided to introduce into the Statistics Act the notion of a voluntary survey. It used not to exist; all surveys used to be mandatory. In 1981 Parliament said there is data that we think we need, but not enough to force people to respond. That potential was introduced into the act, and over the subsequent years, most surveys.... Most household surveys in Statistics Canada are now voluntary.
NDP
Brian Masse NDP Windsor West, ON
But those are surveys. There's a big difference, because they're backed by the census.
NDP
Brian Masse NDP Windsor West, ON
And this is why the United States actually went back to mandatory.
I'd like to ask our other two witnesses this question. There has been a notation that we're going to have a ratio of one to three rather than one to five, but the problem there is the quality of the data, not the volume. I still see some vulnerabilities there.
Also, what are your thoughts about the fact that with the national survey some people--I think--will pick and choose, will cherry-pick it? There will be some questions they are willing to answer and others that they will not be, and I think that's going to require an interesting response in terms of how we statistically deal with that, because it certainly will skew the overall numbers.
But also, even if people pick common things, will that even be statistically meaningful if they're cherry-picking through the national survey? I really think that is what the end result will be. People will fill it out, or a certain amount of it, but they'll get the questions to which they can't be bothered or don't want to actually divulge answers.... That's my concern in this.
Former Chief Statistician of Canada, Statistics Canada, As an Individual
Well, of course we are speculating. One of my main concerns initially about this development was that it wasn't tested and it wasn't discussed. It was just implemented. It required Statistics Canada to implement something. It has never happened in my experience, not in Canada or in any of the developed countries that I know about, that a major change of methodology would have been introduced without testing, without public discussion, and without some very explicit deliberation.
So at this point, since it wasn't tested, we are speculating. My guess would be—but it's a guess, a speculation—that complete non-response would be a bigger problem than a partial response. Once people cross the threshold of deciding to respond, they'll probably respond to most of the questions. That's my guess. But it hasn't been tested, and therefore it's not better than a guess.
Chair, National Statistics Council
Some of this discussion goes to one of the central issues, which is that it is possible to test these things if you organize in advance, and the Americans did this under a very similar circumstance with the American community survey, which is mandatory and is in many ways their analogue to our long-form census.
Somewhat counterintuitively, because had you asked me to guess, I would have agreed with you that what is called “item non-response” would be higher under a voluntary circumstance. They found that not to be the case. “Item non-response” was at about the same level whether it was mandatory or voluntary.
On the other hand, it was also that multi-year and I suppose extremely expensive and detailed research that the U.S. Census Bureau did that showed that non-response bias was a very significant problem. They concluded that, no matter how much effort they put into a voluntary census, it couldn't meet their quality standards because of the non-response bias.
In some ways, to me the lesson is that what is really important when you're about to engage in a major change like this is to do the necessary testing and research in advance before you make your decisions.
Conservative
The Chair Conservative David Sweet
That is pretty well seven minutes right on the button with Mr. McKinnon's answer, Mr. Masse.
Now we'll move on to our second round of five minutes.
Mr. Rota, for five minutes.
Liberal
Anthony Rota Liberal Nipissing—Timiskaming, ON
Thank you very much, Mr. Chair.
I'd like to continue from where we started off with Mr. Smith about the data. It wasn't about the storing of the paper data. It was the storing of the electronic data once it gets into the system. It doesn't stay intact with the owner. The data is strictly that: it's data. It's not assigned.
The concern the minister brought up was that someone would hack in and get to know everything about the individuals who filled out the form. Is that correct?
Chief Statistician, Statistics Canada
We have to retain the records intact, ultimately, so we do have them, and they're in our electronic system. We normally do not store the identifying information with the data itself. We don't need to, but the information is there, and while we're in the process of collecting the information there is always some risk, because we're moving data around the country. You can't say it's riskless, but it is very secure.
Liberal
Anthony Rota Liberal Nipissing—Timiskaming, ON
Has there ever been any break-in or any concern that way? There is always a concern, but has there ever been an event where someone got in--
Chief Statistician, Statistics Canada
No. I'm not aware of any incident of the data being compromised.
Liberal
Anthony Rota Liberal Nipissing—Timiskaming, ON
Very good. I'll just base this on some questions that came earlier, just to clarify.
As it stands right now, Canada doesn't have a general census. That's correct. We just have a survey.
Chief Statistician, Statistics Canada
There is only one census. It is the 2011 census. It has 10 questions and it is a census under the terms of the Statistics Act. There is a census.
Liberal
Anthony Rota Liberal Nipissing—Timiskaming, ON
Is it a survey or is it a census? How do you explain the difference between the two?
Chief Statistician, Statistics Canada
A census by definition usually means a 100% count. Normally, anything involving sampling would not be considered a census in the normal parlance. The provisions in the Statistics Act for a census are quite distinct from those for a survey. A census is automatically mandatory. The questions are determined by the Governor in Council.
Liberal
Anthony Rota Liberal Nipissing—Timiskaming, ON
Then by definition we don't have a census: we have a survey.
Chief Statistician, Statistics Canada
No, we have a census. It is a genuine census. It is going to 100% of the population. It has 10 questions. That will allow us to generate a variety of estimates.
In addition to that, we also have a voluntary survey called the national household survey, which is voluntary, and it asks a series of additional questions--65 in total.
Liberal
Anthony Rota Liberal Nipissing—Timiskaming, ON
Thank you.
Mr. McKinnon, would you like to comment on that?
Chair, National Statistics Council
That's exactly my understanding. A census, in normal statistical jargon as we use it in Canada, means you are contacting everyone, whether it is compulsory or not. What we call our 2011 census, going back to the 1851 census, has happened to be compulsory but they're also universal. What used to be called the short form will be the 2011 census.