The federal government has advanced a number of arguments to justify the abolition of the long-form questionnaire in the 2011 census in favour of a new voluntary "National Household Survey." Each claim is flawed, faulty or incomplete, and together they display a misunderstanding of the nature and purpose of census-taking, and a worrying approach to governing.
Tony Clement has made three assertions or suggestions. The first is that the National Household Survey, to be sent to 33 per cent of
households (the long-form questionnaire goes to just 20 per cent), will be just as accurate.
But while surveys are inherently statistical, the census is primarily a counting exercise.
The census has its own sampling procedures to ensure data accuracy. But those problems can be man-
aged. According to a Statistics Canada technical report on sampling in the 2006 Census, "calibrating sample estimates to known population counts as part of the census weighting procedures helped to reduce the impact of biases."
Voluntary surveys, on the other hand, are known to create much larger biases, with lower participation from the poor, the very rich and aboriginals. These generate results that do not reflect the population and require adjustments.
These biases are harder to detect without a good base of objective knowledge. Increasing the sample size, from 20 to 33 per cent, is irrelevant; cancelling the long-form census removes that objective base.
The long-form census also creates reliable data that is the basis for almost all other major sampling exercises in Canada.
Ivan Fellegi, chief statistician of Canada for 23 years, said that the sampling for the Labour Force Survey, which generates unemployment statistics, is based on long-form census data.
And if the long-form questionnaire is abolished, it will be hard to compare the data to past censuses.
In other words, the new survey cannot use the same standards as past censuses, nor can it deliver the same level of data accuracy, and it compromises future surveys.
Mr. Clement has stated that the long-form census questionnaire has generated many privacy complaints.
Canada's Privacy Commissioner has received just three complaints in the last two censuses. The House of Commons has no record of any petition tabled by an MP about these concerns. Nor has it come up in private members' statements.
In addition, governments still do mandatory "intrusive" surveys. The Labour Force Survey is mandatory. The Canada Revenue Agency requires disclosure of all sources of income.
And completion of the 2011 Census of Agriculture is also mandatory; failure to participate can lead to a fine or prison.
That census requires a report on the area of fields irrigated, the number of live honeybee colonies hosted and the incurred cost of veterinary services, to name just three examples, of every farm in Canada.
Finally, Mr. Clement said that Statistics Canada vouches for this new process.
However, the agency itself refused to go on the record in support, saying, "Statistics Canada is not in a position to answer questions on the advice it gave the Minister in relation to recent statements the Minister has made."
If it fully supports the measures, Statistics Canada should be forthcoming with its explanation and analysis.
Taken together, the government's arguments fail to justify the course chosen.
These are not abstract considerations. The elimination of the long-form questionnaire will result in losses across many areas of Canada's public life, well described by academics and policy-makers.
Yet the federal government has yet to offer any reasons for getting rid of it that can stand up to scrutiny. And that is no basis for public policy of any sort.
Follow us on Twitter: