Guide to the Census of Population, 2021
Chapter 9 – Data quality evaluation
Introduction
This chapter focuses on evaluating the quality of data from the Census of Population. The first section explains why these evaluations are done and how the results are used. The second section lists and describes the main types of possible errors. The third presents the evaluations of coverage of the Census of Population. The fourth section describes data certification.
For the 2021 Census of Population, significant changes were made to the quality indicator dissemination strategy to enable users to do a detailed evaluation of the quality of the data based on their particular needs. These changes, as well as the available quality indicators, are described in Section 5. Section 6 deals with sampling error.
The last sections present information on and measures of the quality of 2021 Census of Population data.
Why evaluate the quality of census data
Census data provide statistical information on housing and on the Canadian population at fine geographic levels and for small subpopulations. These data support planning, administration, development and evaluation of policies by all levels of government. Canadian communities use census data to plan employment, education and health care services. The Census of Population also collects the data needed to update the official population estimates used to determine federal transfer payments to the provinces and territories. Data from the 2021 Census of Population will also be used to determine the number of federal electoral districts and their geographic boundaries, as provided for in the Electoral Boundaries Readjustment Act.
It is essential to ensure the quality of census data. One way to do this is by conducting a variety of evaluations. Quality evaluation activities take place throughout the census process, beginning before data collection and ending after dissemination. These evaluations are based on the six dimensions of data quality presented in the Statistics Canada Quality Guidelines Catalogue no. 12-539-X: relevance, accuracy, timeliness, accessibility, interpretability and coherence. The purpose is to ensure that census data are reliable and that they meet user needs.
Many census data quality evaluations focus on their accuracy; i.e., the extent to which the statistical information describes precisely what it should be measuring. The findings of the activities to evaluate data accuracy are used to validate and certify the data prior to publication, to inform users of the reliability and accuracy of the data, to improve the next census, to adjust census counts for non‑response and, following coverage studies, to produce official population estimates.
Main types of errors
However well a census is designed, the data collected will inevitably contain errors. These errors can occur at virtually any stage of the process, from preparing materials and creating the list of dwellings to collecting and processing the data. Census data users need to be aware of the different types of errors that can occur and of the steps taken to minimize these errors, so that they can evaluate the relevance and accuracy of the data and determine whether they meet their needs.
There are two main types of errors: sampling errors and non-sampling errors. Non-sampling error is likely to bias estimates. Efforts to minimize these errors are made at each stage of collection and processing to reduce their impact; e.g., correcting non‑response and coverage errors by imputing and adjusting the weighting of the data from the long-form questionnaire. However, a residual error remains following this process. Four types of non-sampling errors can occur.
Coverage errors occur when people or dwellings are omitted, counted more than once, or incorrectly counted (i.e., they should not have been counted in the census). Coverage studies are conducted to measure the misclassification of dwellings and the undercoverage or overcoverage of people (see the section entitled Evaluating data coverage, in this chapter).
Non-response errors occur when some or all information about individuals, households or dwellings is not provided. A distinction is made between partial non‑response (no response to one or some questions) and total non‑response (no response to the survey because the household could not be reached or refused to participate).
Response errors occur when a question is misunderstood or a characteristic is misreported by the respondent, the census enumerator, or the Census Help Line operator. They may also occur when data from sources other than traditional collection are used and when the concepts being measured are not exactly the same as those of the survey, or when these data contain errors.
Processing errors can occur at any stage of processing. Responses may be entered incorrectly during data capture, or the coding of responses may be incorrect. Processing errors can also occur during imputation, when a valid response (which is not necessarily accurate) is inserted into a record to replace a missing or invalid response. File manipulation errors are another example of processing errors.
Sampling errors apply only when answers to questions are obtained from a sample. Therefore, this type of error applies only to the Census of Population long-form questionnaire. A sampling error is the difference that would be observed between the estimate from the long-form questionnaire and the true value of the population if there were no non-sampling errors, i.e., all the types of errors mentioned above. It is inevitable when conducting a sample survey, such as the one that uses the census long-form questionnaire (see the section entitled Measuring sampling errors in this chapter).
Evaluating data coverage
Many coverage error studies have been done for recent censuses to help users to assess the impact of coverage errors and to better understand how these errors occur. As part of the 2021 Census, several reviews are designed to improve or evaluate census coverage.
Three studies are conducted to measure coverage errors:
(1) Dwelling Classification Survey
One of the sources of coverage errors in the census is the misclassification of dwellings. This error can occur when an occupied dwelling is classified as unoccupied, or when an unoccupied dwelling is classified as occupied. The purpose of the Dwelling Classification Survey (DCS) is to study these types of classification errors and to adjust counts, if necessary. A sample of dwellings classified as unoccupied or non-respondent is selected, the occupied ones are determined, and information is collected on the number of usual residents for dwellings that are occupied.
This information is used to adjust census data for dwellings, households, and persons. This is done by correcting classification errors and by controlling, using the DCS results, the distribution of the size of households that will be imputed for dwellings that did not return their questionnaire. It is done in time for the initial population count release.
(2) Census Undercoverage Study
The Census Undercoverage Study (CUS) provides an estimate of the number of persons omitted from the census (after taking into account the adjustments described in the DCS above). Estimates are developed for each province and territory and for various population subgroups (e.g., by age, sex and marital status).
- For provinces, the CUS is done in two stages:
- Step 1: Selection of a sample from various sources that encompasses all persons who should have been counted in the census. These sources are the previous census, birth records, immigration and non-permanent resident records, and the sample of persons omitted from CUS of the last census.
- Step 2: Linking persons selected in Step 1 to the Census Response Database (RDB) to determine whether these persons were enumerated. The survey is then used to trace and interview persons who could not be linked with certainty to the RDB to collect additional information. People who died or emigrated before Census Day are identified using administrative data sources, such as death files from vital statistics, during screening or through interviews.
- For the territories, Step 1 involves linking individuals whose names appear in health insurance files with the RDB to identify the ones who were counted during the census. The CUS sample is then selected from the individuals who are not linked, and the remainder of the methodology is similar to the one used for the provinces.
- The CUS results are the major source of information about persons missed by the census. However, unlike the DCS, these estimates are not used to adjust census data before the initial population count release.
(3) Census Overcoverage Study
Double-counting of persons is determined by searching for linked records that have a high degree of matching by sex, date of birth and name. Linked records are sampled and checked manually; the results are used to estimate census overcoverage (or the number of people enumerated more than once). It should be noted that overcoverage caused by the enumeration of persons who were not part of the census target population is not estimated because this component is considered small compared with multiple enumerations.
The combination of the CUS results and the Census Overcoverage Study (COS) results provides an estimate of the net coverage error in the census data. This net error is used to calculate the official population estimates for the Canadian population for each province and territory.
Additional information on the methodology of the DCS, the CUS, and the COS, as well as the detailed results on the coverage of the previous census, can be found in the Coverage Technical Report, Census of Population, 2016, Statistics Canada Catalogue no. 98-303-X. It should be noted that the CUS was referred to as the Reverse Record Check until the 2016 Census.
Certification
Certification consists of several activities to rigorously evaluate the quality of census data at specific levels of geography to ensure that the quality standards for release are met. This evaluation includes the certification of population and dwelling counts, as well as variables related to dwelling and population characteristics.
During certification, a large number of quality measures and indicators are analyzed, such as non‑response rates, invalid responses, edit failure rates, coding accuracy rates, imputation rates and the comparison of data before and after imputation.
The tabulations of the 2021 Census of Population and the estimates from the long-form questionnaire are produced and compared with the corresponding data from previous censuses, other surveys, and various administrative sources. Detailed cross-tabulations are also checked to ensure consistency and accuracy.
An analysis of estimates with outliers is carried out to identify geographic regions with extreme characteristics compared with others and to validate the reasons for these differences with internal and external experts.
Additional checks are also made to minimize the risk that file manipulation errors may have slipped in during data processing.
Various mapping and data visualization tools are used throughout the certification process to make data exploration easier.
Depending on the certification results, census data can be released in various ways:
- The data may be released unconditionally because they are of suitable quality.
- Otherwise, the data may be released conditionally or with restrictions. In this case, they are disseminated and accompanied by a special note warning users of possible limitations. Data can also be processed in a particular way, for example, by combining reporting categories to address quality or confidentiality concerns.
- On rare occasions, data can be deleted for quality reasons. This is the case for incompletely enumerated reserves and settlements (see Appendix 1.5).
For more information on quality indicators and certification results, view the reference guides for each domain of interest on Statistics Canada’s Reference materials, 2021 Census webpage.
Quality indicators
The accuracy of census estimates can be affected by the majority of the aforementioned potential sources of errors. To enable users to conduct a detailed evaluation of the data quality and to determine the relevance of the data to their needs, new quality indicators accompany the 2021 Census of Population data outputs. These include the total non‑response rate and, for each question, the non‑response rates and imputation rates. For long-form questionnaire estimates, which are derived from a sample survey—and therefore subject to sampling error—quality indicators based on variance are also available.
The purpose of the data quality indicators provided is to paint a detailed picture of the risk of known and measurable potential errors at the time of dissemination, such as non-response errors, processing errors, data source errors and sampling errors. These are data accuracy indicators that determine whether the statistical information accurately describes what it should measure.
Users should consult all available quality indicators to ensure that the 2021 Census data meet their needs. More information on quality indicators are provided in the 2021 Census Data Quality Guidelines, Statistics Canada Catalogue no. 98-26-0006.
Total non‑response rate
Total non‑response occurs when all questions are unanswered for a dwelling that received a questionnaire or when a returned questionnaire does not meet the minimum content (i.e., information is not sufficient to continue processing). It is measured by the total non‑response (TNR) rate, which is the primary quality indicator that accompanies each disseminated output from the 2021 Census of Population. In this sense, it replaces the global non‑response rate (GNR), which was used for the 2016 Census of Population and for previous cycles. The GNR combined total and partial non-response, while the TNR rate considers only total non-response. Partial non‑response is considered separately (see quality indicators by question below). This new approach provides detailed information on data quality.
The TNR rate is a measure of non-response that reflects the estimation step. This means that it is calculated by considering corrections to the classification of non-respondent households using the results of the Dwelling Classification Survey. Since all households are enumerated in the census, the TNR rate calculated for data from the short-form questionnaire is not weighted. For long-form data, the TNR rate is weighted to take sampling into account. Therefore, it is an estimate of the proportion of households that would be non-respondent, if all households in the population were interviewed.
Non‑response is a potential source of bias in census counts and in long-form estimates. Bias occurs when the characteristics of respondents differ from those of non-respondents. The TNR rate may indicate the risk that a significant bias has been introduced by non‑response and, where applicable, its potential magnitude. For a given profile of non-respondents, a lower TNR rate indicates a lower risk of non‑response bias and, therefore, more reliable figures and estimates.
To maximize the amount of information disseminated, no data suppression based on non‑response was done for the 2021 Census. However, data for regions with a high TNR rate must be used with caution. A warning about this accompanies data products for which the TNR rate is above 50%.
Comparison between the total non‑response rate in 2021 and the global non‑response rate of previous censuses
The 2021 TNR rate and the GNR of previous censuses meet the same objective: to measure the scope of non‑response in a given region. Conceptually, the difference observed between the GNR of a previous census and the 2021 TNR rate for a given region can be broken down into two parts: the difference due to the change in definition and the actual difference in non‑response rates between the two cycles. The GNR includes partial non‑response and is generally higher than the TNR rate (though it is possible, by definition, for it to be lower). In addition, the GNR is influenced by household size, which is not the case for the TNR rate.
A comparative study of the two indicators conducted on the same dataset showed that their difference is generally less than 5%. Greater differences were observed more often for the indicators on the long-form questionnaire than for those on the short-form questionnaire. When the GNR and the TNR rate are compared, differences of less than 5% can be considered as being solely attributable to the change in definition.
Quality indicators by question
New quality indicators per question were developed for the 2021 Census of Population. They include non‑response rates and imputation rates per question.
The non‑response rate per question is a measure of missing information due to non‑response to a question. The types of non‑response (i.e., total or partial) taken into account by the non‑response rate per question differ for short-form and long-form questions because total non‑response is treated differently for the two types of questionnaires. More specifically, the non‑response rate per question captures only the non‑response that gets resolved through imputation (not reweighting). It can thus be compared to the imputation rate per question described further down. Like the TNR rate, the non-response rate per question is weighted for long-form data. For the same non-respondent profile, a lower non‑response rate per question indicates a lower risk of non‑response bias for estimates derived from a particular question.
The imputation rate per question is used to measure the extent of data processing for each question. Imputation is used to replace missing data in the event of non‑response or when a response is found to be invalid. For this reason, the imputation rate is linked to the non‑response rate, but it also takes into account corrections made to data considered incorrect at the edit stage.
More information on quality indicators per question is provided in the reference guides for each domain of interest on Statistics Canada’s Reference materials, 2021 Census webpage.
Quality indicators based on variance
Since long-form estimates are derived from a sample survey, they are subject to an additional error: sampling error. Variance reflects the variability in estimates due to the use of a sample, not the total population. Sampling variance is therefore estimated using a statistically appropriate method, i.e., one that takes into account the sampling plan and the estimation strategy. The following quality indicators are derived from this estimate of variance.
Standard error
The standard error associated with an estimate is the square root of its estimated variance. A lower standard error indicates a more accurate estimate. Standard error is a key factor in deriving other measures of variability, such as the coefficient of variation, in constructing confidence intervals and in making statistical inferences (e.g., determining whether an estimate is significantly different from a given value or from another estimate).
Confidence interval
The confidence interval was selected as variance-based quality indicator to support the 2021 Census of Population long-form estimates because it helps users easily make a statistical inference.
A confidence interval is associated with a confidence level, generally set at 95%. A 95% confidence interval is an interval constructed around the estimate so that, if the process that generated the sample were repeated many times, the value of the interest parameter in the population would be contained in 95% of these intervals. The usual confidence interval assumes that the sampling distribution of the estimator is a normal distribution. In this case, the lower bound of the 95% confidence interval is estimated by subtracting approximately twice the standard error from the estimate. The upper bound is estimated by adding approximately twice the standard error to the estimate. When the sample size is small, and for certain statistics such as proportions, the assumption that the estimator distribution is normal is violated. Therefore, a confidence interval constructed like this is not appropriate, i.e., its coverage is not 95%.
Consequently, the confidence intervals presented with the 2021 Census of Population long-form estimates are produced using more elaborate methods, which offer coverage closer to the nominal rate. That said, all confidence intervals are based on assumptions that could not be confirmed for some estimates. Further details on the different methods used to construct confidence intervals and their assumptions are provided in the Sampling and Weighting Technical Report, Census of Population, 2021, Statistics Canada Catalogue no. 98-306-X.
Measuring sampling errors
Several factors influence sampling error. Sampling error is smaller when both the sampling fraction and the sample size are large. Ultimately, if the sampling fraction is 100%, as for the census short-form questionnaire, the sampling error will be nil. It will also be small if the variability of the variable of interest in the population is low. This error also depends on the effectiveness of the sample design. For example, it will be smaller if the populations in the strata of the sample design are fairly homogenous or, for a characteristic measured at the person level, if the individuals in the households are rather heterogeneous.
Finally, sampling error depends on the estimation methods used, such as the weighting methods, since some are more effective than others. For example, when the survey weight is adjusted so that a weighted total is equal to the census total, the sampling error for that weighted total is nil. However, it is impossible to adopt a weighting method that would eliminate sampling errors from all possible estimates that could be drawn from the long form.
Sampling error cannot be measured directly. Instead, the actual value of the variable of interest in the population would need to be known to subtract it from the estimate drawn from the long-form questionnaire. This estimate should not include any non‑sampling errors. However, variability measures such as standard error, coefficient of variation, and confidence interval are indicative of the magnitude of this error (see Appendix 1.8).
2021 Census of Population response rate
The response rate is one of the key quality measures used for the Census of Population. Table 9.1 presents the 2021 Census of Population response rates nationally and for each province and territory. Rates are presented for three distinct groups:
- all occupied private and collective dwellings for which a questionnaire was to be received (i.e., response to the short form)
- the subset of occupied private dwellings for which a questionnaire was to be received (i.e., response to the short form)
- occupied private dwellings for which a long form was to be received. For the long form, unweighted response rates and weighted response rates are provided.
Table 9.1 shows the response rates obtained through data processing and data quality assessment. They are calculated as the number of dwellings for which a questionnaire was filled out divided by the number of dwellings classified as occupied according to the census database. The final classification of dwelling occupancy status is based on the analysis of data collected by field staff, data provided by respondents, results of a quality study on the occupancy status of a sample of dwellings (the Dwelling Classification Survey), and administrative data used to impute data on non-responding households in geographic areas with low response rates (see Appendix 1.7).
The response rates in Table 9.1 differ from the 2021 Census of Population collection response rates previously published for occupied private dwellings because they take into account data processing and verification of dwelling occupancy status. These response rates are therefore considered final. Weighted response rates are based on the long form’s sampling weights. Weighted response rates are calculated as the weighted number of sampled private dwellings for which a questionnaire was completed divided by the weighted number of sampled private dwellings classified as occupied.
The response rates shown in the first column of Table 9.1 include collective dwellings. These rates are consistent with the total non‑response (TNR) rates for the short-form questionnaire that accompany released products. The weighted response rates for the long-form questionnaire shown in the last column of Table 9.1 are consistent with the TNR rates for the long-form questionnaire that accompany released products (also weighted).
Table 9.2 presents the TNR rates for the short- and long-form questionnaires nationally and for each province and territory; these rates accompany released products for the 2021 Census of Population.Note 1 Nationally, the TNR rate is 3.1% for the short-form questionnaire and 4.3% for the long-form questionnaire.
| Region | Short-form questionnaire – Occupied private and collective dwellings | Short-form questionnaire – Occupied private dwellings | Long-form questionnaire – Occupied private dwellings (unweighted) | Long-form questionnaire – Occupied private dwellings (weighted) |
|---|---|---|---|---|
| Response rate (%) | ||||
| Canada | 96.9 | 96.9 | 94.9 | 95.7 |
| Newfoundland and Labrador | 96.9 | 97.0 | 95.0 | 95.6 |
| Prince Edward Island | 97.6 | 97.6 | 96.5 | 96.8 |
| Nova Scotia | 97.1 | 97.1 | 95.6 | 96.1 |
| New Brunswick | 96.8 | 96.8 | 94.8 | 95.7 |
| Quebec | 97.1 | 97.1 | 95.7 | 96.3 |
| Ontario | 97.2 | 97.2 | 95.8 | 96.2 |
| Manitoba | 96.5 | 96.5 | 93.1 | 94.4 |
| Saskatchewan | 95.5 | 95.5 | 91.8 | 93.5 |
| Alberta | 96.4 | 96.5 | 93.4 | 94.4 |
| British Columbia | 96.5 | 96.5 | 94.0 | 95.1 |
| Yukon | 95.7 | 95.7 | 85.5 | 89.5 |
| Northwest Territories | 91.8 | 91.8 | 86.2 | 89.2 |
| Nunavut | 79.8 | 79.7 | 78.1 | 78.1 |
| Source: Statistics Canada, Census of Population, 2021. | ||||
| Region | Short-form questionnaire – Occupied private and collective dwellings |
Long-form questionnaire – Occupied private dwellings (weighted) |
|---|---|---|
| Total non-response rate (%) | ||
| Canada | 3.1 | 4.3 |
| Newfoundland and Labrador | 3.1 | 4.4 |
| Prince Edward Island | 2.4 | 3.2 |
| Nova Scotia | 2.9 | 3.9 |
| New Brunswick | 3.2 | 4.3 |
| Quebec | 2.9 | 3.7 |
| Ontario | 2.8 | 3.8 |
| Manitoba | 3.5 | 5.6 |
| Saskatchewan | 4.5 | 6.5 |
| Alberta | 3.6 | 5.6 |
| British Columbia | 3.5 | 4.9 |
| Yukon | 4.3 | 10.5 |
| Northwest Territories | 8.2 | 10.8 |
| Nunavut | 20.2 | 21.9 |
| Source: Statistics Canada, Census of Population, 2021. | ||
Comparability of estimates from the 2021 Census and the 2016 Census programs
Users must be careful when comparing estimates from two censuses or surveys, as they can differ significantly in methodology, quality and target population.
The target population of the 2021 Census short-form and long-form questionnaires was the same as for the 2016 Census. The estimates from the 2021 Census and 2016 Census programs were both derived from mandatory surveys with very high response rates. The response rate for the 2021 Census is 96.9% and for the long-form questionnaire it is 95.7%. These rates are slightly lower than the 2016 response rates of 97.4% and 96.9%, respectively.
Due to slightly lower response rates for the 2021 Census, the non-response error may be greater for some estimates from the 2021 Census program than for estimates from the 2016 Census program. This is particularly true for smaller domains of interest where non‑response may have been greater in 2021 due to the unique collection challenges encountered in Northern or remote regions of the country, and Indigenous communities (see Appendix 1.4). The quality of the estimates and the risk of bias vary slightly more between different communities for the 2021 Census, compared to the 2016 Census.
The quality of the estimates for a given geographic area varies across census cycles based on response rates and incompletely enumerated reserves and settlements (see Appendix 1.5 for information on the increase in incompletely enumerate reserves and settlements in 2021 compared to 2016). When comparing estimates from the 2021 Census and 2016 Census for a given geographic area, users should be mindful of large differences in response rates as well as significant changes in the list of incompletely enumerated reserves and settlements.
Comparisons of estimates from the 2021 Census and the 2016 Census programs for a particular variable can also take into account differences in imputation rates (see previous section on quality indicators by question for more details on this indicator available in both cycles). Table 9.3 presents national-level imputation rates for variables from the 2021 Census and the 2016 Census program, as presented in the 2021 and 2016 topic reference guides.Note 2 Overall, the imputation rates are slightly higher for the 2021 Census short-form questions compared to the 2016 short-form questions, reflecting the lower national response rate in 2021 compared to 2016. For questions asked only on long-form questionnaires, roughly half of the questions have a higher imputation rate in 2021 compared to 2016, while the other half have a lower imputation rate in 2021. Two main factors pulling in opposite directions contribute to this:
- higher total non-response in 2021 compared to 2016 for Northern or remote areas of the country and Indigenous communities (see Appendix 1.4);
- lower rates of missing or invalid information for long-form questionnaires received. This is related to the increased usage of the online questionnaire for the 2021 Census compared to the 2016 Census. Online respondents are prompted when invalid data are entered or if data are missing, which helps reduce the amount of missing or invalid information for each question and improve data quality (see Chapter 6).
In summary, users are encouraged to use all data quality indicators available to judge the quality of estimates from the 2021 Census and the 2016 Census programs when assessing the reliability of comparisons (see the 2021 Census Data Quality Guidelines for more information on data quality indicators). Users are also encouraged to read any quality notes that may be included with dissemination products.
| 2021 Census question or concept | 2021 Census | 2016 Census |
|---|---|---|
| percent | ||
| Question 2—Sex at birth | 3.5 | 2.8 |
| Question 3—Gender | 3.9 | Note ...: not applicable |
| Question 4—Date of birth | 3.7 | 3.1 |
| Question 5—Marital status | 4.7 | 4.3 |
| Question 6—Common-law status | 5.0 | 5.1 |
| Question 7—Relationship to Person 1 | 3.6 | 3.2 |
| Question 8—Knowledge of languages | 4.5 | 4.0 |
| Question 9a—All languages spoken at home | 4.3 | Note ...: not applicable |
| Question 9b—Language spoken most often at home | 4.4 | 3.9 |
| Question 10—Mother tongue | 4.8 | 4.3 |
| Question 11—Canadian military experience | 3.2 | Note ...: not applicable |
| Question 13—Primary or secondary schooling in French in Canada, for residents of Canada outside of Quebec | 5.0 | Note ...: not applicable |
| Question 14—Type of program of schooling in French in Canada, for residents of Canada outside of Quebec | 7.0 | Note ...: not applicable |
| Question 15—Number of years of primary and secondary schooling in a regular French program in a French-language school in Canada, for residents of Canada outside of Quebec | 9.4 | Note ...: not applicable |
| Question 16—Primary or secondary schooling in an English-language school in Canada, for residents of Quebec | 5.4 | Note ...: not applicable |
| Question 17—Number of years of primary and secondary schooling in an English-language school in Canada, for residents of Quebec | 9.4 | Note ...: not applicable |
| Total income derived from the Canada Revenue Agency's tax and benefits records | 5.3 | 4.4 |
| Question 19—Place of birth | 0.9 | 1.0 |
| Question 20—Place of birth of parent 1 | 1.7 | 1.8 |
| Question 20—Place of birth of parent 2 | 2.6 | 1.6 |
| Question 21—Citizenship | 0.8 | 1.3 |
| Immigrant status from Immigration, Refugees and Citizenship Canada’s administrative dataTable 9.3 Note 1 | 2.2 | 0.7 |
| Year of immigration from Immigration, Refugees and Citizenship Canada’s administrative dataTable 9.3 Note 1 | 10.6 | 9.4 |
| Question 23—Ethnic or cultural origin | 8.0 | 4.5 |
| Question 24—Indigenous group | 1.1 | 1.1 |
| Question 25—Population group | 1.6 | 2.0 |
| Question 26—Registered or Treaty Indian status | 1.3 | 1.4 |
| Question 27—Membership in a First Nation or Indian band | 2.1 | 1.8 |
| Question 28—Membership in a Métis organization or Settlement | 7.7 | Note ...: not applicable |
| Question 29—Enrollment under an Inuit land claims agreement | 7.8 | Note ...: not applicable |
| Question 30—Religion | 1.8 | Note ...: not applicable |
| Question 31—Mobility one year ago | 1.5 | 1.8 |
| Question 32—Mobility five years ago | 2.1 | 2.4 |
| Question 33—High school diploma or equivalency certificate | 1.4 | 1.2 |
| Question 34a—Registered apprenticeship or other trades certificate or diploma | 1.7 | 1.8 |
| Question 34b—College, CEGEP or other non-university certificate or diploma | 1.8 | 1.8 |
| Question 34c—University certificate, diploma or degree | 1.6 | 1.4 |
| Question 35—Major field of study | 4.1 | 4.4 |
| Question 36—Location of study | 2.1 | 3.1 |
| Question 37—School attendance | 2.0 | 4.3 |
| Question 38—Hours worked | 1.8 | 1.6 |
| Question 39—On lay-off or absent | 5.7 | 4.5 |
| Question 40—Future start of new job | 3.1 | 4.2 |
| Question 41—Job search | 2.9 | 3.6 |
| Question 42—Availability to work | 2.8 | 3.1 |
| Question 43—Date last worked | 3.5 | 6.2 |
| Questions 44 and 45—Industry | 6.1 | 6.2 |
| Questions 46 and 47—Occupation | 6.7 | 5.3 |
| Question 48—Class of worker | 5.1 | 3.7 |
| Question 49—Incorporation status | 4.1 | 5.1 |
| Question 50a—All languages used at work | 2.8 | Note ...: not applicable |
| Question 50b—Language used most often at work | 2.9 | 3.1 |
| Question 51—Place of work status | 2.9 | 3.7 |
| Question 51—Location of workplace | 3.8 | 5.4 |
| Question 52b—Main mode of commuting | 3.3 | 4.3 |
| Question 52c—Commuting vehicle occupancy | 3.7 | 3.8 |
| Question 53b—Commuting duration | 6.3 | 5.3 |
| Question 54a—Weeks worked during reference year | 4.9 | 2.9 |
| Question 54b—Main reason not working full year | 6.0 | Note ...: not applicable |
| Question 55a—Mostly full-time or part-time work during reference year | 3.3 | 5.4 |
| Question 55b—Main reason working mostly part time | 4.8 | 2.8 |
| Question 56—Amount paid for child care | 3.9 | 4.0 |
| Question 57—Amount paid in support | 3.5 | 4.3 |
| Question 58—Household maintainer | 2.5 | 2.0 |
| Question E1—Tenure | 2.8 | 1.8 |
| Question E2—Condominium status | 2.4 | 1.3 |
| Question E3a—Rooms | 5.0 | 3.6 |
| Question E3b—Bedrooms | 3.2 | 1.8 |
| Question E4—Period of construction | 3.7 | 2.9 |
| Question E5—Condition of dwelling | 2.9 | 1.7 |
| Question E7a—Electricity payment | 6.8 | 6.8 |
| Question E7b—Fuel payment | 6.5 | 7.0 |
| Question E7c—Water and other service payment | 6.7 | 7.0 |
| Question E8a—Rent | 5.4 | 5.4 |
| Question E8b—Subsidy status | 5.5 | 5.1 |
| Question E9a—Mortgage payment | 5.0 | 5.1 |
| Question E9b—Property taxes included in mortgage | 4.5 | 4.1 |
| Question E9c—Property taxes | 7.4 | 7.4 |
| Question E9d—Value of dwelling | 7.1 | 7.1 |
| Question E9e—Condominium fee | 15.3 | 14.4 |
| Question E10—Monthly use or occupancy payment for dwelling | 44.8 | Note ...: not applicable |
... not applicable
|
||
Comparability of the variability of estimates from the 2021 and 2016 Census long-form questionnaires
As mentioned in the previous sections, estimates produced using data from a sample survey, such as those from the 2021 Census long-form questionnaire, include sampling error, i.e., the error stemming from the fact that only a sample of the population was observed. Sampling error is determined using variability measures such as standard error or the coefficient of variation (CV). In Appendix 1.8, CVs are used to compare the variability of estimates from the 2021 and 2016 Census long-form questionnaires.
Moreover, the purpose of the 2021 and 2016 Census long-form questionnaires was to produce estimates for a series of questions for a variety of geographic areas, ranging from very large areas (such as provinces and census metropolitan areas) to very small areas (such as neighbourhoods and municipalities), and for various population groups, such as Indigenous peoples and immigrants. These groups also vary in size, especially when cross-classified by geographic area. Such groupings are generally referred to as "domains of interest." The purpose of this section and of Appendix 1.8 is to compare the variability of estimates from 2021 and 2016, not to compare the estimates. However, sampling variability should be taken into account when comparing estimates from these surveys, particularly for small "domains of interest," since the observed differences can be the result of sampling variability rather than an actual difference in the population.
- Date modified: