Wednesday, August 11, 2010

'Fixing' what isn't Broken: Tampering with the Census

I have been startled this last few weeks to witness the extent to which a seemingly esoteric subject, the census, has become a political and conversational issue during the dog days of summer. Here in Calgary I have had a series of conversations about this topic, and so far all of them have been with people upset by the changes. Most people have been concerned by the prospect of impeding the gathering of good data, and everyone I have had the conversation with have been baffled by the government's decision to fix something that manifestly wasn't broken.

The decision by the Conservative government of Mr. Harper to change the long-form census from mandatory to optional was announced and defended on two grounds. First, that the current system was intrusive, and thus needed to be changed on privacy grounds. Second, that because there were legal sanctions attached to a failure to return the mandatory form it was necessary to act in order to protect Canadians from being punished by the state should they decide not to return the form.

Both of these grounds were, sadly, completely fatuous. In the case of the latter it is enough to point out that no Canadian has ever been jailed for the failure to complete the census. If there is a case I've missed on this point please let me know! As for the former, well, there are a series of reasons why that concern makes little or no sense. To begin with it is entirely possible to return the form blank, or with inaccurate answers. Tens of thousands do - witness the write-in answer of 'Jedi' under the religion category over the past few censuses. Secondly anything which identifies your answers with you is held in confidence for 92 years, at which point you will be past caring about what you may have said. Thirdly the data you are submitting via the long form frequently exists elsewhere, in tax filings, building permits, school records etc, all of which are vastly more accessible than the census data. Fourthly, many of us submit vast amounts of personal data to private organizations, whether those be stores or banks, that maintain and are held to a much lower standard of privacy. Finally it should be borne in mind that in an era in which we all can and do complain about things which irritate us every day the last census received a grand total of three, yes three, privacy complaints. I would submit that any private organization that sent out a questionnaire to its membership and received only three complaints would be thrilled; and that on a vastly smaller sample size.

The problems with the proposed new system of a mandatory short-form and a voluntary long-form are manifold. To begin with the mandatory long-form is the fundamental control group for all of the research done in this country by Statistics Canada and most private researchers as well. The mandatory long-form is invaluable because it reaches a huge number of people, the response rate is high and the people are randomly chosen. As a result the data provides a statistical sample size and randomness that is unmatched. This means that its results are accurate even though many people either don't complete it or send back that they are Jedi - these outliers fall off the edges of the curve for any given question and we are left with an enormous sample to work with. the result is data that is internationally admired for its reach, accuracy and completeness. To make the sample voluntary means that response rates fall, and the response rate skews away from a representative sample of Canadians. The worst thing about this is that we won't be able to know how much, or which way, the data has skewed - we will have cancelled the control group we test everything else against.

What I have found fascinating the past few weeks is the number of people I have had this conversation with, many of whom are not political people, much less policy wonks. I have several times been told that we, as Canadians, prefer to settle our differences by appealing to the facts. There seems to be something in the census changes that people recognize as an attack on the facts, or at least the system for gathering them. Last time I checked there were hundreds of groups, ranging from provinces and municipalities to professional groups like statisticians and economists, in opposition to the changes. In favour these was only the government, the National Citizen's Coalition (formerly run by Mr. Harper) and the Fraser Institute. The contrast is educational in and of itself. The issue also seems to have resonated with a much wider section of the public than the narrow element of the Conservative Party's base that the move was intended to satisfy. the polling data the last week indicates that it has hurt the government's standing with voters, though like all polls (especially in the summer) that needs to be placed in context. Unless the opposition is able to make strides towards portraying itself as a viable government-in-waiting then the results are unimportant.

Mr. Clement's announcement today that the government would move the language section of the long-form to the short-form census questionnaire is a transparent effort to avoid the court challenge filed by the Federation of Francophone and Acadian Communities of Canada. The conflict between the maintenance and extension of the short-form census and the rhetoric about removing the mandatory long-form and protecting citizen privacy from the state is striking.