On August 17, Statistics Canada presented in The Daily the main results of the 2021 census with respect to language. It deals with knowledge of official languages, mother tongues and languages spoken at home. For some censuses, the languages spoken have been the subject of two questions, or sections: all persons surveyed are assigned at least “one language spoken most often”, while seeking to count, where applicable, the persons speaking less “another language regularly”. We are interested in these two aspects.
The Daily informs us that in 2021, “85.5% of the Quebec population declared that they speak French at home at least regularly” and that ” [p] Almost 1 in 5 people in Quebec (19.2%) spoke English at home at least regularly”. Thus, even before considering Quebecers who speak a third language at least regularly, we are already at 104.7%! The addition of third languages, or 15.6%, brings the total to 120.3%!
Faced with this astonishing total well above 100%, the question that must be asked is as follows: first, where does this 1.7 million people come from who pushes the total beyond 8 .4 million counted in Quebec in 2021; then, what do the results mean?
Statistics Canada justifies this way of doing things in a “guide”. It is a booklet of more than 40 pages entitled Interpretation and presentation of census language data. We have made a critical review of it published in The duty from 1er last March. Statistics Canada proposes to perform two types of inadmissible additions in descriptive statistics.
On the one hand, it is recommended to add the data according to the language spoken most often with those relating to the language spoken regularly, which goes against number theory (a Venn diagram would show the doubling of languages spoken regularly). On the other hand, it is suggested to count twice the double answers (of the type “French and English”) and three times the triple answers (“French, English, other”).
One would look in vain in the manuals of social statistics for a justification of this way of proceeding. Nothing better than a very small example to show the inadmissibility of these two ways of doing things. If among 1000 owners of a principal residence there were 100 also owning a secondary residence, the addition of these two numbers would take us from 1000 owners to 1100 properties. Also, if most of these properties were made of wood or brick, no one would suggest double counting those with wood AND brick siding.
According to the STATCAN “guide”, ” [p] o take the measure of the presence of a language, it [serait] preferable to take into account all mentions of the language in question”. To take into account all the mentions of a language is to make a “sum of occurrences” of the answers collected. However, this type of sum leads to an indicator of little interest.
Indeed, in such sums, all mentions have the same value. On the one hand, there is no longer any difference between the languages spoken “most often” and those spoken “regularly”, because they are placed on an equal footing. On the other hand, since two languages are mentioned by the bilinguals, their answers are added to the two languages in question. As for trilinguals, their numbers also increase each of the three languages mentioned.
The attachment of some analysts from the same unit at Statistics Canada to this kind of exercise comes, in my opinion, from an excessive desire to present the demographic evolution of Francophones in Canada in a way that is less dramatic than a neutral reading of the data. For example, in the case of Quebec, while the distribution of data on the language spoken most often shows a regression from 80.6% to 79.1% between 2016 and 2021, the addition of all the mentions shows a loss , yes, but at a higher level: from 87.1% to 85.5%.
Elsewhere in Canada, New Brunswick is similar to Quebec in this respect. While the demographic weight of our francophone cousins in this province has slipped from 28.6% to 27.0% in five years, the addition of all the mentions maintains this weight above 30%, despite a decline of 33.2% in 2016, down to 30.4% in 2021.