Murtaza Haider is a professor of management at Ryerson University. He is currently working on a book on data science and analytics, which will be available in spring 2015.
Independent reviews of the National Household Survey (NHS) data have revealed its inferior and dubious quality, which has led to calls by researchers to avoid it altogether. With $22-million more spent on the NHS than on the long-form census, and many more individuals surveyed than in the past, the NHS has failed to achieve the objectives of its proponents and at the same time, it has delivered data of limited use.
An opportunity exists for legislators to right this wrong. A private member’s bill, which calls for the restoration of the long-form census, is up for debate on Thursday (Nov. 6). It is incumbent on the MPs not to think and act along party lines and instead to act in the best interest of Canadians, which lies in the restoration of the long-form census.
Ted Hsu, a Liberal MP from Kingston, has brought forward a private member’s Bill C-626. Being a former research scientist with a Ph.D. in Physics, Mr. Hsu understands the importance of good data and its contribution to research, public policy and business strategy. However, private member’s bills have seldom become law in Canada. He would need the support of not just the Liberals and the NDP, but also the Conservative MPs who hold a majority in the House of Commons.
Bill C-626 makes the following three principal recommendations: First, it calls for reinstating the mandatory long-form census; second, it recommends that the Chief Statistician, in consultation with the relevant stakeholders, should formulate the policies regarding census, including the means of collecting the data; and, third, the bill proposes a consultative process, which engages experts and political parties, for appointing the chief statistician.
Unless the amendments proposed in Bill C-626 are implemented, the status quo does not bode well for developing evidence-based public policy or business strategy in Canada. Already, the researchers at the University of Toronto have exposed the poor quality of NHS income data. A comparison between the NHS income data for 2010 and the one from the Canada Revenue Agency revealed significant and systematic differences. The researchers concluded that the income data “should not be used or cited. It should be withdrawn.”
Even more troubling is the fact that the non-response rates for the NHS are significantly higher than that for the long-form census in the past. According to Sam Boshra, a former analyst with Statistics Canada, if the NHS was subject to the 2006 threshold for quality, the data for 75 per cent of the dissemination areas across Canada would not have been released.
This implies that while a significantly larger number of individuals were contacted for the NHS than for the long-form census in the past, a large proportion of those who were contacted refused to complete the survey. Effectively, we have a self-selection bias. Because the NHS is a voluntary survey, we are unaware of the motivations of those who responded and those who did not. Thus, we are ignorant of our ignorance or what Donald Rumsfeld famously called the “unknown unknowns.”
The self-selection bias, which now plagues all of the NHS data, cannot be rectified. We do not have the means and measures to correct the non-response biases across Canada. Furthermore, in the absence of a reliable census, a large number of other surveys in the public and private sectors are rendered effectively useless because they rely on the long-form census for their sampling frame.
In the age of big data and analytics, the decision to abandon the long-form census has put us behind our global competition, which is relying on data driven analytics to create new products and services.
Bill C-626 provides Canada an opportunity to right that wrong. The MPs should put Canada’s welfare first when they review and vote on the Bill. Voting along the party lines is not in Canada’s interest.Report Typo/Error