The 2021 census unduly disadvantages French

One of the fundamental principles guiding the decisions of census managers is the concern to preserve the comparability of data from one census to another.

calvin veltman
Professor (retired), ESG-UQAM, sociolinguist, University of Quebec in Montreal

Statistics Canada applies this logic, for example, by maintaining a question that invites the most assimilated respondents not to declare their true mother tongue, on the pretext that they have “forgotten” it. Following this logic, my father was Dutch-speaking until my grandmother died, after which he would be a native English speaker. It may not be logical, but it is comparable from one census to another.

However, Statistics Canada has departed from this well-established tradition by modifying the questions on the language of use in 2021. In an article recently published in The Conversation¹I showed that the 2021 census data diverged considerably from those produced by the 2001 to 2016 censuses, in particular by reducing the reporting of second languages ​​regularly spoken at home.

Having said that, we could also ask ourselves whether the new census does not deviate a little too much from the expected values. For example, does it favor French compared to previous censuses?

By the term “expected values”, it should be understood that Quebec society resembles a large ocean liner. Two factors prevent the liner from drifting too far from its trajectory: firstly, the vast majority of the population of 2016 was still alive in 2021. Secondly, the components of the evolution of the population change only very little, even often not at all, over such a short period of time. So we can imagine what the data for 2021 should look like, projecting from the data from 2001 to 2016.

Methodology

First, our treatment of multiple responses recorded in the census departs somewhat from the methodology traditionally applied. Briefly, the person who declares an allophone language, in whole or in part, is assigned to the allophone group for the mother tongue while for the language of use, it is assigned to the host language group.². This method results in a slight increase in the share of host languages, while decreasing the share of non-native languages ​​by the same amount.

Second, the analysis of data from 2001 to 2016 demonstrated the regular evolution of the different components of the population, according to the language of use:

• the regression of French without English (second language) from 78.7% to 75.1%;
• the growth of French with English from 4.2% to 5.7%;
• the growth of French/English bilingualism from 1% to 1.6%;
• the growth of English with French from 2.1% to 2.5%;
• the regression of English without French from 8.1% to 7.9%;
• the growth of allophone languages, with or without second languages, from 5.9% to 7.3%.

Given the regularity of the evolution of these six linguistic groups, we carried out simple regression analyzes on basic data, slightly adjusted to respect the size of the population actually enumerated in 2021. Other methods would undoubtedly give similar estimates.

The results

Now we will present the expected and observed values ​​in 2021.

First, let’s look at the losers: French with English as a second language (regularly spoken) loses more than 161,000 people compared to anticipated values ​​and English with French as a second language, 25,000 people. It is obvious that the new questionnaire discriminates against the declaration of regularly spoken second languages. As for the winning groups, English without French adds more than 81,000 people; French without English, nearly 59,000 and French/English bilingualism, 33,000. Analysis of the estimated gains already shows that the 2021 census favors the English language.

Evolution of indicators

To fully understand the extent of the change, it is necessary to use a few indicators:

• the share of French, the main language, by distributing French/English bilinguals on a pro rata basis, as is customary;
• the share of English, calculated in a similar way;
• the RFA index (the ratio of French to English), ie the share of French in the sum of the two previous measurements.

This table clearly shows the steady decline of French from 2001 to 2016 as the predominant language, from 83.4% in 2001 to 81.6% in 2016 – the anticipated value for 2021 was 80.9%, or 1 percentage point of more than the observed value. As for English, its share rose gradually from 2001 to 2016 and should be around 11.3% in 2021, not 12.2%. As for the indicator of the balance of power of the French, it should be around 87.7%. However, it is only 86.8%, a regression of 1%³.

Disadvantaged French

The data analysis clearly shows how much French is disadvantaged by the new method of data collection. English is favored. No one knows which data is more accurate, those from 2001 to 2016 or those of the last census. To better dissect this question, Statistics Canada could create and make public a microdata file presenting, among other things, all the linguistic variables for each person residing in Quebec during the last two censuses. This would allow the scientific community to hold its own debate on the quality of the 2021 census – unless Statistics Canada decides to return in 2026 with the questionnaire used from 2001 to 2016.

³ If we attribute the second languages ​​spoken in allophone households, French would be disadvantaged by 1.4 percentage points and English, advantaged by 1.1 points.