Next August, Statistics Canada will release data from the 2021 census on language use at home and in the workplace. It would then be preferable for the federal agency not to repeat an error that dates back to the 2001 census.
In Canada, census language data is generally grouped into three distinct classes: French, English and all other languages. In the booklet Interpretation and presentation of census language data (August 2020), the Center for Ethnocultural, Language and Immigration Statistics (CSELI) states that there would be “no need for groups [linguistiques] be defined in a mutually exclusive manner.
Announcing that languages can be classified into non-exclusive groups is another way of saying that some data might be counted more than once. According to the editor, Mr. Jean-François Lepage, an analyst at the CSELI, we should not be surprised to find additions that “exceed[nt] the total population”, i.e. more than 100%.
This booklet puts forward the idea that, in order to “take the full measure of the presence of a language, it is often preferable to take into account all the mentions of the language in question”. However, to count “all the mentions”, Statistics Canada proposes to carry out inadmissible additions in descriptive statistics.
Although it is rightly admitted that “the language [parlée le plus souvent] should weigh more than the tongue [parlée régulièrement] it is wrongly suggested to add each of the categories established according to the language spoken most often with those formed according to the language spoken regularly. The same applies to the languages used at work.
Also wrongly, it is proposed to double the three categories of double answers: 1) French and English, 2) French and a third language, 3) English and a third language. Finally, we suggest tripling the counts of the single category of triple responses in which cases mentioning French, English and a third language are grouped together. As a result, across Canada in 2016, the sum of languages in use reached 120% in homes and 116% in workplaces.
It is important to clearly understand the nature of the results obtained. In reviewing the calculations described by Statistics Canada in The Daily (November 29, 2017), we had identified “sums of occurrences” for French, English and all third languages in response to the two questions on the language of work (Quebec demography notebooks, fall 2019). Very rudimentary, the sums of occurrences do not recognize any hierarchy between the numbers.
Rare are the situations where data is grouped into “non-mutually exclusive” classes, that is, classes that overlap. This is the case of knowledge of third languages. Since respondents are asked to mention ” [toute(s)] language(s) other than French or English”, we agree with the CSELI that it would be necessary to “multiply the categories of multiple responses exponentially” to exhaust all the possible situations.
Thus, the data on the knowledge of third languages does not add up, because the sum of people who know, for example, Spanish on the one hand and those claiming to know Italian on the other hand would have the consequence of counting two times anyone who knows these two languages.
According to Statistics Canada, forming “non-mutually exclusive” classes would result from a new “emerging” approach. This approach would, it seems, be better suited than the classic approach to the study of indigenous minorities, for example. However, the CSELI would have gone it alone in the matter, because nowhere in federal institutions can we find a directorate that would have counted certain people twice or three times simply because they declared so many citizenships during a census.
To “report on the richness, complexity and diversity of linguistic behaviors and situations in Canada” as Statistics Canada would like, the elementary tools of descriptive statistics are sufficient. Neglected in this publication, “crosstabs” have not become obsolete.
On the contrary, the absence of such a table deprives the reader of several interesting pieces of information: 1) in 2016 across Canada, in households where people spoke French most often, 9% of residents also regular use of English; 2) in contrast, only 2.6% of respondents from Anglophone households also spoke French on a regular basis; 3) among those who spoke a third language most often at home, English dominated French as a regularly spoken language by more than 9 to 1.
Comments or suggestions for Ideas in Review? write to [email protected].