Technical report on changes for the 2021 Census
Ethnic or cultural origins: Technical report on changes for the 2021 Census

1. Introduction

The census is the country’s storyteller. Through good and difficult times, the census has been providing a detailed picture of Canada’s changing mosaic—giving Canadians facts on where we came from, to where we are, and guiding decisions on where we want to be. To respond to the country’s needs and write the next chapter of Canadians’ data foundation, Statistics Canada conducted country-wide engagements. We also conducted a census test with over 135,000 households to determine whether new or revised questions under consideration can be easily understood and correctly answered. We listened to Canadians, because we know that this is about your census, your community, your future.

This series of technical reports provides an in-depth view of the changes in content for the 2021 Census of Population, as the census process is ever changing to reflect changes in the Canadian population and its demographic characteristics.

A question on ethnic or cultural origins has been included on the Canadian Census of Population since 1871. Furthermore, even before Confederation, a question on this topic was asked in censuses, reflecting a long-standing, continuing and widespread demand for information about the ethnic and cultural diversity of the population. 

Today, this information is used to support the Canadian Multiculturalism Act and the Canadian Charter of Rights and Freedoms, as well as by governments, businesses, academics, researchers, and ethnic and cultural associations throughout the country.

Responses to the ethnic or cultural origins question on the census reflect respondents’ perceptions of their background. As such, many factors can influence changes in responses over time, including the contemporary social environment, the respondent’s knowledge of their family history, and their understanding of and views on the topic.

Over time, there have also been differences in the wording, format, examples and instructions of the ethnic origin question used in the census. As a result, the historical comparability of ethnic origin data has been affected by these factors, as well as by changes in data processing, proxy responses and the social environment at the time of the census.

2. Approach for the 2021 Census

The content of the 2021 Census questionnaire is the result of extensive engagement, discussions and testing, including comprehensive one-on-one interviews and a rigorous 2019 Census Test involving over 135,000 households across Canada.

After the 2016 Census, concerns were raised that changes to the list of examples of ethnic and cultural origins included as part of the question were affecting response patterns. Concerns were also raised about the wordiness of the question, which made it difficult for certain people to read and respond to the question.

To investigate the impact of examples on response patterns, Statistics Canada analyzed the 2016 Census data integrated with data from the 2011 National Household Survey and found evidence that the examples in the question itself have a significant prompt effect.^{Note 1} Specifically, respondents were more likely to report an origin when it was included in the list of examples and, conversely, less likely to report an origin if it was not included in the list. This analysis was supported by the results of the 2019 Census Test, which are described in the following sections.

To address the limitations of the question asked in 2016 and to better reflect Canada’s growing diversity, examples will be removed from the 2021 Census questionnaire to eliminate their effect on responses.^{Note 2} In their place—and to help respondents better understand the question—a description of types of origins will be provided, along with a link to a list of over 500 examples of ethnic and cultural origins. This extensive list of examples will provide greater detail and diversity than what was disseminated in the past.

3. 2019 Census Test: Ethnic or cultural origins questions tested

The 2019 Census Test included three versions: a control version, which was identical to the 2016 Census question on ethnic or cultural origins, as well as two different test versions.^{Note 3}

The first test version, test version 1, included the following elements:

• There were no examples of ethnic or cultural origins in the question itself.
• Instead, the question included the following instruction: “Ancestors may have Indigenous origins, or origins that refer to different countries, or other origins that may not refer to different countries.”
• Examples were replaced by a link to a page with an extensive list of examples of ethnic and cultural origins. Examples in this list were the result of stakeholder and expert engagement.

The second test version, test version 2, used the same approach as test version 1, but included 10 examples of different types of origins in the question itself (Indigenous origins, origins referring to countries, origins not referring to countries). The purpose of testing this version was to see whether including just a few examples of different types of origins would be enough to help respondents answer the question without introducing the same example-led prompt effect seen in the 2016 Census question, which included 28 examples.

4. 2019 Census Test: Summary of results^{Note 4}

4.1 Data quality

As indicated, the results of the census test supported the findings of the recent technical report^{Note 5} on changes in response patterns related to the ethnic or cultural origins question. In short, including examples of ethnic or cultural origins directly on the questionnaire prompts respondents to report the origins included among the examples rather than those that are not included.

Chart 1 presents the proportion of 2019 Census Test respondents who reported selected origins by version. In the chart, example-introduced response bias is evident in the response patterns for the origins included as examples directly on one or more of the question versions. The origin “Canadian,” which appeared as an example only on the control version, received—proportionately—more than twice as many responses for the control version than for either of the two test versions. Example-introduced bias was also seen for origins such as Punjabi, Jewish and Acadian, which were included as examples only on test version 2 and all had much higher counts for this version than for either the control version or test version 1.

The origin “Scottish” was included as an example on both the control version and test version 2, and there was evidence of higher example-introduced bias in test version 2, where Scottish was 1 of just 10 examples, rather than 1 of 28 examples. As a result, test version 2 should not be viewed as an acceptable middle ground between the control version, which included 28 examples, and test version 1, which had no examples. Instead, the 2019 Census Test showed evidence of even greater example-introduced bias for examples included in a shorter list than in a longer one. Therefore, in this regard, test version 2 had the lowest data quality of the three versions tested.

Test version 1 also yielded more varied and diverse responses. Almost 43% of respondents to test version 1 reported origins that were not among the 28 origins used as examples on the control version. In contrast, 34% of respondents in the control panel reported origins that were not among the 28 examples.

This diversity in response patterns is also seen in the reporting of different categories of origins—Indigenous origins (e.g., Cree, Inuit), origins referring to countries (e.g., Canadian, Chinese) and other origins that may not refer to countries (e.g., Acadian, Jewish). For the control version, almost 90% of respondents reported country-based origins, compared with under 80% for test version 1. This is understandable, as many of the origins used as examples for the control version refer to countries. 

In contrast, for test version 1, a higher percentage of respondents (20.7%) reported other (non-country) origins. For the control version, 10.6% of respondents reported other (non-country) origins.

4.2 Other considerations

Without examples directly on the questionnaire, the ethnic or cultural origins question is for many respondents more difficult to respond to. Based on the results of the 2019 Census Test, this was evident both in the amount of time it took respondents to complete different versions of the question and in the non-response rates for the different versions of the question.

Compared with the control version, both test versions of the question took longer for respondents to complete for their households. This was largely a result of the extra time that respondents spent on the linked list of examples. Without examples directly on the questionnaire, more respondents accessed the linked list of examples, and reviewing the extensive list of examples resulted in longer time needed to respond to the question.

As well, the test versions had higher non-response rates than the control version (control: 2.8%, test version 1: 4.8%, test version 2: 4.1%). Again, this is understandable—having examples directly on the questionnaire helps respondents answer the question, albeit by prompting them to report one or more of the origins used as examples on the questionnaire. Without examples on the questionnaire, more respondents have difficulty answering the question, resulting in higher non-response.

4.3 Summary

In the 2019 Census Test, test version 1 of the ethnic or cultural origins question exhibited the highest data quality of the three versions tested—it was the version that largely mitigated the data quality issue of example-introduced response bias. It was also a more challenging version of the question for many respondents to respond to, as evident in the longer average time it took for respondents to answer the question, and in the higher non-response rate.

However, improving the data quality of the question by removing example-introduced response bias is of primary importance, and for this reason, test version 1 of the ethnic or cultural origins question will be used for the 2021 Census. This revised version of the ethnic or cultural origins question also yielded more varied and diverse responses. More respondents reported origins that were not on the list of examples on the 2016 questionnaire, and more respondents reported origins that were not associated with countries.

5. Conclusion

The 2021 Census question on ethnic or cultural origins has been revised since the 2016 Census and now includes the following elements:

• There are no examples of ethnic or cultural origins listed in the question itself.
• There is a description of different types of origins, specifically: “Ancestors may have Indigenous origins, or origins that refer to different countries, or other origins that may not refer to different countries.”
• There is a link to a page with over 500 different examples of ethnic and cultural origins.^{Note 6}

The revised question addresses concerns regarding the impact the list of examples on the questionnaire was having on response patterns. The question allows respondents to report their origins without the influence of examples listed on the questionnaire, resulting in better-quality data that are more representative of the population.

However, the new version of the question for the 2021 Census will produce results that will not be comparable to the 2016 Census results for many ethnic and cultural origins. The origins that will be particularly affected are those that were among the 28 examples listed directly on the 2016 Census questionnaire (e.g., Canadian). For the 2021 Census, these origins will be included as part of the much more extensive list of examples of ethnic and cultural origins, mitigating the prompt effect they have had in the past.

It should be noted that the results on ethnic or cultural origins have not been fully comparable between censuses in the past either, in part because of changes to the list of examples over time.^{Note 7} After the examples are removed from the questionnaire, results will no longer be affected by this factor, and this will help improve historical comparability moving forward.

The approach for the 2021 Census will yield more varied and diverse responses than in past censuses. To better reflect the range of responses received, a greater number of origins will be disseminated for the 2021 Census.

Appendix 1 – 2019 Census Test questions on ethnic or cultural origins

Appendix 2 – 2021 Census linked list of examples of ethnic and cultural origins (preliminary)