Industry Committee on July 27th, 2010

[Dr. Ivan Fellegi, Former employee of Statistics Canada, As an Individual]

Thank you.

Thank you for inviting me to appear before this committee. I prepared slightly longer opening remarks, which I'm not going to read. They have been distributed or are available for distribution.

I'd also like to mention that I asked the Clerk of the Committee to distribute the National Statistics Council statement, “Seeking Solutions”, which was released yesterday.

I'll restrict my oral comments to a very few issues.

First of all, I want to underline how impressed I am—even I am impressed, having spent 51 years in Statistics Canada—by the widespread support and testament to the usefulness of the census and its importance. I wouldn't have believed that there would be so many groups mobilized and ready to make statements emphasizing their basic, fundamental dependence on the census. There cannot be any dispute about that at this point.

Second, I want to underline that any voluntary survey is intrinsically biased. That matters a great deal, because bias, unlike sampling error, cannot be estimated from the survey data themselves. Sampling error we can estimate. We make statements, as pollsters do, such as that 19 out of 20 times the results are within plus or minus 2% or 3% or whatever. Bias is a sneaky, pervasive risk, which we seek to understand always, but we never do. And the more widespread and the more detailed the breakdown of the data, as is the case with the census, the more pernicious the risk of bias becomes, because we don't know where it crops up. If it's used widely without an appreciation of the likelihood of bias, that's a major societal risk. It's not a statistical risk; it's a societal risk.

You don't have to take my word for it. The Statistical Society of Canada and the American Statistical Association both came out and made totally unambiguous statements on this issue.

What makes the bias particularly worrisome in this context is the fact that most users are really not interested in a snapshot; they are interested in how things have changed since the last time they were measured. And if the last time they were measured they were measured in an unbiased manner, and the next time they are measured in a biased manner, the results become basically not usable for that purpose. Even if they could in some sense be used as a kind of general guide for what the score is now, they really become unusable for purposes of making comparisons in terms of what has happened since the last census. And a great deal has happened since the last census. Among other things, there was a financial crisis in 2008.

My third point that I want to emphasize is that privacy is a central concern at Statistics Canada and has always been, and we have taken innumerable steps to improve the situation that was already very good to begin with.

First of all, there isn't a single case of any information ever reported to Statistics Canada having been released with identification of the source--that is, who reported it.

Second, way back in 1971—and this is just a factual correction I want to make—it wasn't the first time that we used the long form; in 1971 it was the first time we used a short form. Prior to 1971, every census was a long form--complete, 100%.

In 1971, at my personal initiative, actually—I wasn't chief statistician yet, but I was senior enough that I could take that initiative—we introduced sampling into the census for the very first time and created a short form, and the long form went to one in five people.

I won't go into all the other things we have done to improve privacy, because the chair has indicated that my time is maybe up, but I do want to close with really the fundamental point where I hope I'll be questioned—that is, where do we go from here? Statistics Canada and, through Statistics Canada, all its users are facing a unique situation, and we need to find constructive solutions.

I want to call your attention to the solution proposed by the National Statistics Council, which is a body actually appointed by the minister to advise the chief statistician. They came out yesterday with what I consider to be totally constructive proposals for resolving the impasse.

I want to close on that point. Thank you.

[The Chair, Conservative, Michael Chong]

Thank you, Mr. Fellegi.

We'll have about 15 minutes of questions and comments from members, beginning with Mr. McTeague.

[Dan McTeague, Liberal, Pickering—Scarborough East, ON]

Mr. Chair, thank you.

Mr. Sheikh and Mr. Fellegi, I want to thank both of you for being here today. I think it's extremely important. For an issue to have risen its head in the dead of summer, as it has now, is amazing. I've not seen this in my 17 years as a member of Parliament where a decision by the government has raised such opposition throughout the land.

Arthur Carty, the national science adviser; Jean-Pierre Kingsley, the former chief electoral officer; Linda Keen, the nuclear safety commissioner; Peter Tinsley, the former military complaints commissioner; Paul Kennedy, the former RCMP public complaints commissioner; Adrian Measner, former chair of the Wheat Board; and now you, Mr. Sheikh: after 38 years in a distinguished career of public servitude, for which we are indeed grateful, you have now found yourself in a position where you have had to resign.

Mr. Sheikh, could you please inform this committee about your unprecedented and courageous action to resign as chief statistician for this country?

[Dr. Munir Sheikh]

If I tried to write down in detail the answer to the question you have posed, I don't think I would do a better job than the answer posted by Mr. Alex Himelfarb, a previous Clerk of the Privy Council.

With your permission, Chair, I'll just try to read that answer in response to the question. He said, and I'm just going to quote the whole thing from him verbatim, the following:

In Canada, our professional, non-partisan public service has traditionally been guided by the principle of “fearless advice and loyal implementation.” This is based on the belief that governments work best when they have access to the best possible information, options, and advice—including what they may not wish to hear—and, in the end, democracy demands that the public service implement loyally whatever lawful decision the elected government of the day makes—whether the public servants agree or not. That's how it works when it works. I know Munir to be a man of great integrity, committed to the value and values of a professional, non-partisan public service....

Munir also had another responsibility as Canada's Chief Statistician and that is to protect the integrity and credibility of the agency and its products without which Statistics Canada cannot deliver its mandate. In Munir's introduction to the Statistics Canada website he spoke with evident pride of the objectivity and neutrality of the agency's data. Statistics Canada, which has earned a sterling international reputation, has long understood that it can do its job of informing public and private decisions and supporting democratic accountability only if people trust in its integrity and technical competence.

I have not yet had the opportunity to talk with Munir but I imagine that this is why he felt it necessary, when doubts arose about what Statistics Canada advised, to acknowledge publicly that the voluntary approach he was to implement is not a substitute for the mandatory survey. No Chief Statistician would want people to lose trust, to think that Statistics Canada compromised its technical advice to the government or would, in any way, misrepresent the information it provides to Canadians. This goes to the heart of the agency's credibility and of the values of public service.

[Dan McTeague, Liberal, Pickering—Scarborough East, ON]

Mr. Sheikh, I am among many who believe that your integrity, sir, is very much intact, and we thank you for that. But I want to ask you, what led to the resignation? Was it that the minister had forced you, out of the question of principle and integrity, to do something such that you, as an independent arm's-length civil servant of some 38 years, decided that the future of the census at Statistics Canada was very much at stake? Why were you forced to resign, sir?

[Dr. Munir Sheikh]

Let me first of all say that it is the right of the government to make decisions that, if lawful, should be, as I said, loyally implemented by any department of the government.

The reason I resigned, which I made clear in my resignation statement, was that when doubt began to be expressed about the nature of the advice we gave, which to any statistician would come across as not the work of a statistician, I came to the conclusion that I cannot be the head of an agency whose reputation has suffered.

[Dan McTeague, Liberal, Pickering—Scarborough East, ON]

The interference, then, by the government is well founded...the government itself. Listening to the minister's testimony, I'm surprised, sir, that you survived as long as you did.

Mr. Fellegi, there was a comment by the minister in the previous testimony in which he pointed out that the reasons for eliminating the long-form census—not all of them, but some of them—were because of people's perception of giving and contributing to information that might lead them to fear deportation, that people coming from “hideous regimes” might be averse to answering those kinds of detailed questions.

Mr. Fellegi, you come from a totalitarian regime. How well founded are those comments by the minister?

[Dr. Ivan Fellegi]

Well, I grew up in a regime that certainly was totalitarian. It was Hungary in its socialist period. I chose to leave for that reason, as soon as the first opportunity presented itself.

But yes, the answer is that of course it's not founded, unambiguously not founded. Statistics Canada goes to endless lengths to protect the confidentiality of the information it receives. There is no single instance where the Statistics Act had to be used against a person because he or she released such information, knowingly or unknowingly. So it's not just a question of wilful release; wilful or inadequate attention...to make sure that it doesn't happen.

[The Chair, Conservative, Michael Chong]

Thank you, Mr. Fellegi, and thank you, Mr. McTeague.

Monsieur Bouchard.

[Robert Bouchard, Bloc, Chicoutimi—Le Fjord, QC]

Thank you, Mr. Chair.

Good morning, gentlemen.

Mr. Sheikh, I am very happy and pleased that you have agreed to appear before the committee. I would like to congratulate you on your courage and decision to resign. Your decision is of great concern to me.

Mr. Fellegi, I would like to thank you for speaking to the media. Your comments were very insightful. Thank you.

My questions are for Mr. Sheikh in particular.

In your view, what is the main disadvantage of the minister's decision?

[Dr. Munir Sheikh]

Let me clarify a question, one that may be on some people's minds, by giving an analogy. We of course have voluntary surveys, lots of voluntary surveys; we have mandatory surveys; and we have a census. There could be a question as to why you would need a mandatory census if you have voluntary surveys.

I would say that we are like an auto manufacturer who produces passenger cars, SUVs, and 56-footers. We stand behind each one of our products. However, I would never recommend that somebody use a passenger car to deliver merchandise on a regular basis to a Wal-Mart, and I would not recommend to anybody to use a 56-footer with two people inside it to drive around the city of Ottawa. These things are produced for specialized purposes, and as long as they're used for that purpose, they are really great.

The difficulty with a voluntary survey as a replacement for a mandatory census is that, as my predecessor has explained, it does not capture the biases or deal with them. That would not be the case with a mandatory census. The reason we have those problems is that we know for a fact that there are certain geographical areas that would not provide the response rate one would want, and there are certain classes of individuals who have very low response rates. Indeed, these are exactly the people from whom you would want information if the government is to develop policies to deal with their issues. This would include aboriginals, people with low incomes, people with less education, visible minorities, and immigrants. Their response rates are quite low.

So a voluntary survey creates the problem that we will not get information from these groups, which of course will skew the averages in the overall research.

The other difficulty in a voluntary survey of a regular type is that when we get biases we can correct for those biases by comparing the information with the information available in a census. But if the census itself becomes voluntary, then there's no benchmark that one can use to actually fix the problem for the voluntary survey. That is the difficulty we would face.

[Robert Bouchard, Bloc, Chicoutimi—Le Fjord, QC]

Thank you.

You have already answered my second question to a certain extent, but could you expand on the people and groups concerned? In your opinion, are there people or groups who will be more at a disadvantage because of the government's decision?

[Dr. Munir Sheikh]

I don't think it is my place to comment on that, because we are not a policy department. Our job it to basically collect data, and the thing that's on our minds is the quality of data. I think I can go as far as saying that the data are going to be of a lesser quality if it is voluntary survey-based.

How policies will deal with that, or whether it has any impact on policy development, is not an issue on which I am knowledgeable, given the nature of my work. You are probably better off asking policy departments for the answer to that question.

[Robert Bouchard, Bloc, Chicoutimi—Le Fjord, QC]

Would you like to answer this question, Mr. Fellegi?

[Dr. Ivan Fellegi]

I completely agree with Mr. Sheikh on our role.

However, I would be ready to provide you with examples where, as a statistician, I could question the quality of data, especially data on new immigrants and their economic integration in Canadian society, or on aboriginal people. The census is almost the only source of information on their level of education that allows us to find out whether the gap between the Canadian population and aboriginals, as a subset of the population, is growing or shrinking.

Those are just two examples, but there are others where, as a statistician, I questioned the quality of data.

[The Chair, Conservative, Michael Chong]

Thank you, Mr. Bouchard.

Mr. Lukiwski, you have the floor.

[Tom Lukiwski, Conservative, Regina—Lumsden—Lake Centre, SK]

Thank you very much, Mr. Chair.

Thank you both, Mr. Sheikh and Mr. Fellegi, for appearing here.

We have a lot of ground to cover, so I'm going to try to keep my questions short, and I would appreciate if you could try to keep your answers short as well.

Would it be a fair assessment, gentlemen, for me to say that in your role as statisticians, it really doesn't matter to you how the information is collected as long as you get the information you need, but that your position is that you believe, in the census, a mandatory requirement is the most effective way to get the information you request? Would that be a fair statement?

[Dr. Munir Sheikh]

As I said earlier, there are some needs that users have out there for which you need a census. There are other needs that can be satisfied quite easily by a voluntary survey.

The questions that are in the census are, in my view, the questions that can be answered...or most of them are. Maybe there is the odd one here or there that doesn't belong there, but in terms of the long form overall, you're not going to get the reliable information you seek on those questions from a voluntary survey.

[Tom Lukiwski, Conservative, Regina—Lumsden—Lake Centre, SK]

Mr. Fellegi has stated that you believe that the information received from voluntary surveys or census is biased, although, having said that, is it not true that in Statistics Canada, between census, perhaps the majority of your surveys are of a voluntary nature? I speak of things like the Canadian community health survey, the adult education and training survey, and the aboriginal peoples survey.

They are all voluntary, are they not? And if they are, do you utilize the information that you glean from those surveys?

[Dr. Ivan Fellegi]

Yes, they are voluntary, and they benefit from the terrific advantage of having a benchmark like the census against which they can be compared. So we know exactly, when we design the survey and when the data become available, how they compare on those variables that are included in both surveys--how they compare with the census and whether or not some potentially very big biases have crept in that need to be corrected.

[Tom Lukiwski, Conservative, Regina—Lumsden—Lake Centre, SK]

But if I may interject, on their own merit, as a stand-alone voluntary survey, would you make the statement that you believe the information you glean from that is biased?

[Dr. Ivan Fellegi]

If we didn't have the census to compare them against, I would never know the answer. That's what makes it pernicious.

[Tom Lukiwski, Conservative, Regina—Lumsden—Lake Centre, SK]

Do you also agree, sir, that...? I want to get one thing clarified, if I can, on an earlier statement that you had made.

You had said, in response to a question from Mr. McTeague, that you did not believe that...I think you put it “unambiguously”; you did not believe there was any danger of an immigrant feeling threatened by the mandatory nature of the census.

But how can you say that, sir? If there is a threat of imprisonment, or threat of a fine, for failure to answer the census, how can you say with clarity that no one would feel threatened by that?

[Dr. Ivan Fellegi]

I hope I didn't say that. I certainly didn't intend to say that. What I intended to say unambiguously--and I repeat, unambiguously--was that there are no grounds for such fears, because nobody's information has ever actually been released with an identification or with any way through which that information could be traced back to that person.

That was the context of the question. As to whether or not people coming from oppressive regimes have reason to fear when they give information to the census, the answer is that they have no reason to fear.

[Tom Lukiwski, Conservative, Regina—Lumsden—Lake Centre, SK]

But there is a threat that they could be imprisoned or fined if they don't answer the questions, so how can you say that they may not feel threatened?