Use of administrative data to impute non-responding households in areas with low response rates for the 2021 Census

Notwithstanding a very successful response rate of 98% for the 2021 Census, there remain localized areas of the country with lower-than-planned response rates, due to wildfires, COVID and other factors. The dwellings in these areas represent 0.08% of all occupied dwellings in Canada.

Efforts to produce high-quality Census data do not end in the field. Our efforts continue during data processing until we have an accurate picture that reflects all the people of Canada. To produce the best possible information for lower levels of geography, Statistics Canada is drawing on administrative data to support imputation for areas and households where it is deemed appropriate. Nothing substitutes for the voice of Canadians, but administrative data help us ensure every Canadian is included in the portrait of our country produced by the census.

Imputation is an internationally-recognized technique used by many statistical agencies to assign a value to missing information in surveys. Statistics Canada research over the past several years has shown that using administrative data to support imputation can improve the quality of data in some circumstances. Statistics Canada has been drawing on administrative data for many years to supplement missing content to ensure the high quality that Canadians expect of the census. For example, when the city of Fort McMurray was evacuated due to a wildfire during the 2016 Census, Statistics Canada used administrative data to complete the census in the city and surrounding region.

Administrative data will be used for the 2021 Census, during data processing, to impute (assign values to missing data) non-responding households in areas with low response rates, and where the administrative data are of sufficient quality to generate reliable population counts. The administrative data used for imputation of non-responding households are data already provided to the government and meet the highest quality, privacy, security, and confidentiality standards. Learn more about administrative data.

Frequently asked questions (FAQs)

• Given the COVID-19 pandemic and its related restrictions, how did you make sure every Canadian was counted to produce high-quality census data?

  The 2021 Census of Population achieved a successful overall collection response rate, while keeping Canadians safe during the COVID-19 pandemic.

  Statistics Canada increased the number of census responses collected using our online electronic questionnaire (EQ), and actively promoted EQ as the best response method, to all Canadians. For Canadians without internet access, or who preferred to use other safe options, Statistics Canada provided a paper questionnaire or collected the census over the phone. When field follow-up was required, strict health and safety protocols were followed.

• What is a statistical contingency plan?

  Nothing substitutes for the voice of Canadians to ensure every Canadian is included in the portrait of our country produced by the Census. The statistical contingency plan helps us ensure that we provide the most accurate and detailed picture of Canada and our communities using internationally accepted methods.

  Specifically, there can be exceptional situations in a census cycle when data cannot be collected using traditional enumeration methods. For instance, in 2016, wildfires interrupted Census collection in Fort McMurray, Alberta, and as a result, questionnaires were obtained from fewer than 30% of households. In that situation, administrative data was used as a statistical contingency, to fill gaps in the data due to the evacuation.

  Despite a very successful response rate for the 2021 Census (98%), there remain localized areas of the country with lower-than-planned response rates, due to wildfires, COVID-19 and other factors. About 12,000 dwellings (out of a total of 14.8 million occupied dwellings in Canada), or 0.08% of all occupied dwellings in the country, were / will be imputed using administrative data.

• Why do we use administrative data during data processing?

  Nothing substitutes for the voice of Canadians, but administrative data help us ensure every Canadian is included in the portrait of our country produced by the Census.

  Efforts to produce high-quality data do not end after the census collection campaign is complete. Our efforts continue during data processing until we have the most accurate possible picture that reflects all the people of Canada. To produce the best possible information for lower levels of geography, Statistics Canada is drawing on administrative data to support imputation for areas and households where it is deemed appropriate.

• What is the definition of imputation?

  Imputation is a recognized technique used by statistical agencies around the world to assign a value to missing information in surveys. Based on other available data and historical patterns, imputation allows for information from a responding household to be assigned to a household which has not responded in whole or in part to the survey, improving the quality of the resulting statistical information.

  Imputation eliminates gaps in the data and, when done appropriately, reduces bias introduced by non-response. This is traditionally done by identifying persons or households that have characteristics deemed similar to the incomplete record (the “donor pool”) and by copying (“imputing”) their information to fill in the missing responses.

• What does Statistics Canada consider a safe and appropriate use of administrative data?

  The safe use of administrative data means that Statistics Canada places the highest priority on the privacy, security, and confidentiality of the administrative data sources it uses.

  The administrative data used for the imputation of non-responding households are data already provided to government. Before we acquire any new administrative data, the proposed acquisition goes through a rigorous “necessity and proportionality” evaluation, a privacy impact assessment, a data ethics review, and a legal review to ensure it complies with the Statistics Act. When we receive new administrative data, we “anonymize” the data as soon as possible by removing personal identifiers and replacing them with anonymous ID numbers. Finally, only Statistics Canada employees can access administrative datasets, on a need-to-know basis.

  The appropriate use of administrative data means that the quality of the administrative data for a certain geographic area, community, or household has been assessed, and Statistics Canada has determined that administrative data could be appropriately used to assist imputation for non-respondent households after collection.

• Under what circumstances would Statistics Canada implement the imputation plan?

  Statistics Canada will use administrative data to assist the imputation of non-responding households under the following circumstances:

  • in areas where response rates are below 90% due to wildfires, COVID-19 and other factors
  • only for dwellings where good quality administrative data are available
• Does Statistics Canada have experience using administrative data to impute non-responding households?

  Yes. During the 2016 Census, Fort McMurray, Alberta was evacuated due to a wildfire. Less than one-third of the population were able to complete the census. However, Statistics Canada was able to quickly design and implement a contingency plan using administrative data, and was able to generate an accurate population count for the community.

  Since 2016, Statistics Canada has further developed its methods and ability to use administrative data under exceptional circumstances, and is prepared to use it for imputing data to non-responding households after collection. This internationally accepted approach helps to ensure the production of high quality data that Canadians expect of the Census.

• What census characteristics could Statistics Canada measure using administrative data, if a non-responding household is imputed?

  For census purposes, the administrative data used for imputation of non-responding households could be used to determine the number of usual residents of a household, their date of birth, and their sex at birth. Other short-form variables (such as language, gender, census family characteristics) will be imputed using the “traditional” imputation method, which is an internationally established and recognized method that can produce valid population-level estimates.

• What are the administrative data files included in the imputation plan?

  The administrative data that would be used in the imputation plan are data already provided to federal and provincial government agencies. These data sources are all government-sourced and come primarily from the Canada Revenue Agency, Immigration, Refugees and Citizenship Canada, and Vital Statistics (births, deaths), and driver's license files made available from the provinces and territories.