Sunday, March 04, 2012

The former Chief Statistician of Canada speaks

On February 28, I had the pleasure of listening to talk by Munir Sheikh, the former Chief Statistician of Canada, entitled, "The Importance of Evidence-Based Policy Development". In it, he described his tenure at Statistics Canada, the government's decision to make completion of the long form optional, and the reasons for his resignation from Statistics Canada.

Dr. Sheikh compared the validity of the 2006 census to the 2011 "survey" (he refuses to consider an optional long form a census). When given a survey, about half of Canadian households will complete it without further prompting. If there is a reminder, such as a follow-up telephone call, then compliance usually increases to about 70%. But only when compliance is mandated by law do participation rates increase to over 99%, as the Canadian census enjoyed until 2006.

But perhaps, Dr. Sheikh asked rhetorically, seventy percent is sufficient to draw important conclusions about the demographics of Canadian society. Unfortunately, response rates differ drastically across different groups. Recent immigrants to Canada, aboriginals, and those with low levels of education are far less likely to complete the long-form survey. Conversely, those with high income levels disproportionately participate voluntarily. Thus the results from the 2011 survey are inherently skewed, but the extent of the bias is impossible to determine.

This concept was explained to the government, which chose to proceed with eliminating the mandatory element of the long-form questionnaire contrary to the advice of Dr. Sheikh and Statistics Canada. The stated reason was concern for the privacy of Canadians. Dr. Sheikh revealed that out of a population of over 35 million Canadians, there were a total of two complaints about the 2006 census related to privacy concerns.

Yet the reason Dr. Sheikh resigned from public service was not that the government decided on a course of action against his advice. His duty as a civil servant was to provide recommendations based on the public interest but ultimately to follow policy as set by the federal government. It was only when the government claimed that Dr. Sheikh counseled that "a voluntary survey can become a substitute for a mandatory census" with equally reliable results that he submitted his letter of resignation. "It can not," he wrote.

His speech was followed by a question and answer session. Many people wanted to know (or speculated themselves) what the government's true motivation was for eliminating the long-form census. The audience seemed convinced it was a mixture of Conservative party ideology and a firm commitment to avoid the possibility of policy being questioned by facts. Dr. Sheikh consistently refused to speculate on motivations, sticking strictly to the facts and refusing provide further commentary. His professionalism was exemplary. It is possible that Canada could return to a proper census in 2016, he said, if the government so chose. However, the Conservative party announced its intention to eliminate the census before the 2011 election, and achieved its first majority. Therefore the likelihood of an about-face on this issue is unlikely.

I had the opportunity to ask two questions. If one wanted demographic data for any reason in 2013 (after all the results from the 2011 survey are published), what source would be best to use? The 2006 census, the 2011 survey data, or some combination of the two? Were there reliable sources other than Statistics Canada?

 Dr. Sheikh responded that it would be best to use the 2006 census as a baseline, and to use the 2011 survey to inform about trends since then. However, good judgment would need to be used when using any data from the 2011 survey, since there is no means to determine how accurate its results are. Other surveys are not particularly helpful since, by definition, they are completed by those that fill out questionnaires.

I also asked about the census question on religion, which is asked on the long form every second census (once a decade). It asks about a person's religion, with instructions to "Indicate a specific denomination or religion even if this person is not currently a practising member of that group."

This question, as worded, measures the faith group one grew up with, not how one presently identifies oneself. If one is non-practicing, then "No religion" is probably the most accurate descriptor - but an unknown (presumably large, given the discrepancy between the 2001 census and recent survey data) number of non-practicing individuals will probably select from the list of denominations provided. Given his knowledge of extensive experience, does Dr. Sheikh believe it likely that Statistics Canada will improve the wording of this question in the future?

Dr. Sheikh replied that Statistics Canada was aware of the problem, and that it had met with groups defending each side of the issue. It will also likely consider the question again prior to the 2021 census. But there is tremendous value in maintaining the same question over long periods of time, to allow for comparison and analysis over the decades. Therefore there needs to be something fundamentally wrong with a question (or the improvement needs to be vast) in order for Statistics Canada to change its wording. My impression was that Dr. Sheikh believes it is unlikely the wording of the question on religion will be changed.

Overall, I was very impressed with Dr. Sheikh's intelligence, discretion, and articulation. The federal civil service lost a consummate professional when he resigned in 2010.

No comments:

Post a Comment