Save the census

Journal name:
Nature
Volume:
466,
Page:
532
Date published:
DOI:
doi:10.1038/466532a
Published online

The Canadian government should rethink its decision to change the way census data are collected.

It is hard enough to get people excited about statistics at any time, let alone at the height of the summer holiday season. But in Canada this month, people have become passionate about the subject. A quiet political decision to scrap the compulsory long-form part of the Canadian census has inspired fuming in the national press, vitriolic protest from numerous academic bodies, the resignation of the head of Statistics Canada, and even a YouTube broadcast of a song espousing the importance of census data for public policy (http://go.nature.com/O9TWcf). Critics have accused Canada's ruling conservative party of being anti-science — and, worse, anti-information. To gut one of the world's most respected statistical organizations, they argue, will prevent future policy decisions on everything from health care to public transport from being based on either data or logic.

The incident comes amid a growing sense of unease about the right-leaning Canadian federal government's apparent disregard for science-based policy. The country continues to support the mining of asbestos and its export to the developing world, despite repeated calls to ban the toxic substance and cries of protest from the medical community. Canada has been one of the most obstructive countries at climate-change talks, and continues to be protective of its development of the tar sands — one of the world's dirtiest sources of oil. The federal government has fought against maintaining the supervised injection facility for drug addicts in Vancouver, despite staunch protest from the medical community and studies showing that such programmes are helpful. Now the government is threatening to undermine the system that collects the data needed for a multitude of other evidence-based decisions.

Every five years for almost four decades, the Canadian census has involved both a head count and a longer questionnaire asking about details such as ethnicity, education and housing. The long form was sent to 20% of the population, and replying was compulsory by law. At the end of June, citing concerns about privacy, industry minister Tony Clement announced that the long form would be made voluntary; an additional Can$30 million (US$29 million) a year will be spent to send the form to 30% of the population, in an attempt to make up the numbers.

“Vulnerable populations of the poor and downtrodden will be less likely to reply to a voluntary questionnaire, skewing the resulting data.”

This, as any statistician can testify, is not the same thing. Inevitably, vulnerable populations of the poor and downtrodden will be less likely to reply, skewing the resulting data. Although statisticians are adept at correcting for such factors in surveys, they can do so only if they have a gold-standard set of data to refer to — namely, the census data. Even if the voluntary data are sufficiently robust to allow for good public policy-making, it will still cause a problem for researchers looking for long-term trends, because comparing the compulsory data with the subsequent voluntary set will be nearly impossible.

In a country defined by a mosaic of immigrant groups and indigenous populations, precision data on their lives are crucial to good public policy. Statistics Canada has been widely regarded as a world leader in handling everything from the intricacies of question wording to protecting individual privacy. Although the government claims that it is responding to public concerns about privacy and 'big brother' government, this, ironically, does not seem to be based on evidence — the privacy commissioner's office has received only three complaints about the intrusiveness of the census long form over the past decade.

The United States investigated the option of switching to a voluntary long form in 2003, but concluded that it would be too expensive to bring the data up to par. Instead, this year it replaced its long form with a mandatory ongoing survey — a change that required nearly a decade of research to confirm that it would maintain the quality of the data. Academics, economists and public policy-makers find themselves open-jawed at Canada's snap decision — made without consulting the data's users — which will effectively allow the government to do less while spending more money. Letters of protest and editorials have been penned by the Canadian Medical Association Journal, the Canadian Federation for the Humanities and Social Sciences, and many others.

Canadian academics are not letting this pass quietly; the rest of the world should join their voice of protest (http://go.nature.com/dq6rZ3). It is too late to save the form for 2011, but this could, and should, be reinstated for 2016.

Comments

  1. Report this comment #12507

    Michael Yeo said:

    ?Science-based policy? is a remarkably vague idea, and given to sloganistic use even. The idea that scientific evidence relevant to policy decisions should be sought or produced and taken into account in policy decision-making is clear enough, and compelling. However, the phrase lends itself to being used in other less clear and compelling senses, especially as deployed in rhetorical contexts when the one using it is advocating for a particular policy outcome in a particular case.

    One danger, and it occurs quite frequently in public discourse, is of a conceptual slippage from: a) policy that respectfully takes into account scientific evidence as this evidence is relevant to the policy decision; b) policy that is supposedly directed by scientific evidence (as if values were not also an essential component of policy-decision-making), and that selects the policy option preferred or recommended by those who are expert in the relevant science (in effect both obscuring, and privileging, the values that the experts, in some instances quite self-interestedly, happen to have).

    The danger here has been referred to in the literature as ?generalization of expertise?. It is necessary, and quite reasonable, to defer to experts in the matters in which they are expert. However, this deference, and such authority as experts may have qua experts, is sometimes transferred to experts (or assumed by them) even when their opinion goes beyond expert opinion and is a value judgment. For example, the opinion of a statistics expert that a voluntary census will yield data significantly flawed and inferior compared to the mandatory long form census is properly an expert opinion. A lay person who argued with the statistician on this point would be foolish and a policy maker who neglected to take it into account would be missing an essential piece. However, the opinion of the statistical expert that we should therefore maintain the mandatory long census is not an expert decision but a value decision, a properly political decision that should take into account not just the scientific evidence (and of course ideally the best evidence available) relevant to the policy issue, but also the values engaged in and by it. The science alone does not and cannot yield policy decisions; for that, values are also necessary.

    Whatever your opinion about the census issue, this opinion, on one side of the policy issue or the other, will be determined ultimately by your values. And that is true also for the experts who advocate for the policy that the long-term census should be maintained. Their expertise in the scientific evidence relevant to this question does not give them any special authority to pronounce on the policy issue. Expertise in the evidence does not carry over into expertise in the policy question (there is no ?expertise? in the policy question as such ? it is not a matter for experts).

    I think such a generalization of expertise occurs in the following passage from the editorial:

    ?The incident comes amid a growing sense of unease about the right-leaning Canadian federal government's apparent disregard for science-based policy. The country continues to support the mining of asbestos and its export to the developing world, despite repeated calls to ban the toxic substance and cries of protest from the medical community. Canada has been one of the most obstructive countries at climate-change talks, and continues to be protective of its development of the tar sands ? one of the world's dirtiest sources of oil. The federal government has fought against maintaining the supervised injection facility for drug addicts in Vancouver, despite staunch protest from the medical community and studies showing that such programmes are helpful.?

    When the ?medical community? calls for this or that policy option regarding the policy options mentioned, this is not an expert judgement properly speaking. For example: ?What are the effects of asbestos?? is an expert question; ?should its import be banned? is a policy question and in making it it will be quite appropriate to weigh the negative health impacts against other considerations. The opinion of the medical community about this ban (their value judgement) should not be confused with their opinion about harmful effects, etc.

    The cause of science-based policy will best be served if: a) politicians leave the expert questions to the experts and ensure that expertise/evidence is brought to bear on policy decision-making as appropriate; b) experts do not generalize their expertise when commenting on policy issues and usurp the role of politicians, or obscure the value issue by representing it as if it were a scientific one.

Subscribe to comments

Additional data