The long-form census

Another day, another column (here) on the defunct long-form census.  I am a social scientist, I use a lot of data, and a lot of census data (albeit from 200 years ago) – but I am not pining for us to return to pre-2010 days when Canadian HAD to fill out the long-form census.  And I don’t think arguments for a return to the pre-2010 status quo are all that compelling.  Why so?

  1.  Prior to 2010 the penalty for not complying with the long form census was 3 months in prison + $500 fine.  To my mind, it’s simply outrageous that we would imprison – or even threaten to imprison – people for refusing to answer questions such as how much unpaid housework they had done in the last week. The argument that such harsh penalties were never actually sought cuts no ice with me: i) the Crown sought exactly that in this case; ii) there ought not to be laws on the books that the government can opportunistically resurrect – it breeds contempt for the law and invites capricious government.  I’m hardly on the fringe in this respect: a) when the government got rid of the long-form census all 3 opposition parties conceded that a 3-month sentence was over the top and ought to be repealed; b)Ted Hsu’s bill (C-626) itself conceded the same point, and reduced the fine for refusal to $500 (see Section 8).
  2. The reason that some folks clamour for the old LFC is that the harsh sentence for non-cooperation essentially resolved the problem of (non-random) self-selection into the survey.  (Although as Maioni’s column points out, compliance was still only at 93.5%.)  Non-random selection poses a real problem for statistical inference. Let’s understand, however, that C-626 would not have resolved that issue, it would simply have changed its nature:  prior to 2010 only the 6.5% of people who were presumably willing to risk 3 months in jail and pay $500 did not respond; if C-626 had passed, some % (> 6.5, I wager) of  people willing pay $500 would not respond.  And it’s not quite so simple as saying that less non-response is always better; it depends on the random (or not) nature of that non-response: a survey that 50% of respondents answer or not on a truly random basis may be  better than one which 80% of respondents answer or not on a non-random basis.  I submit – though it’s a matter of pure conjecture – that it’s the latter situation that Ted Hsu’s bill would have left us in. (And no, this is not a reason to bring back the threat of a 3 month prison term!)
  3. I accept the need for a census, and that there is a compelling public interest in having the state provided with certain pieces of information, mainly the number and age of its citizens.  I’m sure a good case can be made for furnishing the state with other details. However, I can see no compelling public interest for knowing how much unpaid housework people are doing, whether my house requires minor repairs (Question H5), even the age of my house (which the municipality already knows).  I think that a lot of otherwise sensible people don’t seem to get that in an ostensibly liberal society it’s the state that must make a compelling case to interfere in the citizen’s life.  Each and every question in the LFC ought to have been subject to this sort of acid test.  This was and is not the case:  Sec. 22 (especially sub-clause u) of the Statistics Act pretty much empowers the Minister to obtain statistics (and hence compel response – see Sec. 31) on virtually anything — quite opposite to the general principle that the onus ought to rest on the government.
  4. Does Stats Can seriously think that they were not getting massive measurement errors in some of these questions?  Really, what differentiates a major from a minor house repair?  No criteria are offered (e.g., is it a structural vs a non-structural matter, or is it about the cost, anything over $10,000 for example?).  These sorts of questions generate variables that are almost useless when put on the RHS of an equation; errors-in-variables in spades IMO. And for this we threaten people with prison and fines! Ay carrumba.

I think Stats Canada would be better off figuring out how to make its NHS more effective OR cull through the LFC questions, decide what information is really critical, and then make a reasoned and compelling case to the Minister to gather that information in the short-form census.

Leave a Reply

Your email address will not be published. Required fields are marked *