Coverage Technical Report, Census of Population, 2021
1. Estimates of population coverage errors

1.1 Introduction

The census defines the population to be counted and the rules by which it is to be counted (see Section 3). Coverage errors occur when errors are made relative to these definitions and rules. The main sources of coverage errors include failure to include a dwelling (and, in turn, failure to include its residents) and respondent error (not including all persons who should be included or including persons who should not be included). This section presents estimates of 2021 Census population net undercoverage, undercoverage and overcoverage. Undercoverage and overcoverage may lead to bias in official counts and estimates because the characteristics of the persons not included may differ from those of persons who are included, and the characteristics of those counted more than once (duplicates) may differ from those of persons who were counted only once. Net undercoverage indicates the extent to which the number of enumerations included in census data is higher or lower than complete enumeration.

1.2 Net undercoverage

The 2021 Census population net undercoverage rate was estimated at 3.00%.^{Note 1} In other words, the difference between the number of persons who were not included in the census but were members of the census target population and the number of duplicates was estimated to be 3.00% of the census target population. The population undercoverage rate is estimated at 4.98% (1,897,876 persons), while the population overcoverage rate is estimated at 1.98% (755,635 persons). An undercoverage rate of 4.98% indicates that persons who were not included but who were part of the target population represent 4.98% of the census target population. An overcoverage rate of 1.98% indicates that duplicate enumerations represent 1.98% of the census target population.

The estimated undercoverage rate has increased compared with the 2016 Census, while the overcoverage rate has been relatively stable. As a result, the net undercoverage rate, which is the difference between undercoverage and overcoverage, increased in comparison with the 2016 Census.

Since the primary goal of census coverage studies is always to produce the best coverage estimates possible for the most recent census, their methodology for the 2021 Census, as described in Section 7 and Section 8, has been improved. However, these improvements limit the comparability of these estimates with the 2016 net undercoverage estimates and partly explain the difference observed. In addition, since the net undercoverage estimates are measured from samples, the margin of error associated with the 2016 and 2021 estimates may also partly explain the difference.

This section presents estimates of net undercoverage based on the following geographic and demographic variables:

• province or territory of current residence at the time of the census
• age and gender
• marital status and gender
• mother tongue
• census metropolitan area (CMA) of the usual place of residence on Census Day.

Table 1.2b provides an estimate of net undercoverage, standard errors related to the estimate, and the corresponding estimated net undercoverage rate and standard error for various characteristics. Negative net undercoverage estimates indicate that the overcoverage rate was higher than the undercoverage rate. For an explanation of how this can occur, see Section 9.

The standard error provides an indication of the accuracy of sampling-based estimates. An interval covering two standard errors on both sides of the estimate includes the correct value 19 times out of 20. In other words, there are approximately 19 chances out of 20 (95%) that the actual population net undercoverage rate for the 2021 Census was between 2.8% and 3.2% (i.e., 3.0% ± two standard errors) or about 2 chances out of 3 (68%) that the actual rate was between 2.9% and 3.1% (i.e., 3.0% ± one standard error).

Since net undercoverage reflects both undercoverage and overcoverage, readers should also consult the undercoverage and overcoverage estimates in Table 1.3. For example, a low net undercoverage rate may reflect a low undercoverage rate, or a high undercoverage rate combined with a high overcoverage rate.

The highest population net undercoverage rates in the country were observed in the three territories: the Northwest Territories and Nunavut had the highest rates (8.04% and 7.82%, respectively), followed by Yukon (5.78%). Among the provinces, Prince Edward Island had the highest net undercoverage rate, at 4.28%, followed by British Columbia (4.08%) and Ontario (3.95%). Quebec had the lowest population net undercoverage rate, at 0.50%, followed by New Brunswick (1.73%). All other provinces had a net undercoverage rate that was at least one percentage point higher than that of New Brunswick. In 2016, the highest and lowest rates were also observed in the Northwest Territories and Quebec.

Beginning with the 2021 Census, the estimates of census coverage error are reported by gender rather than by sex. Gender refers to an individual’s personal and social identity as a man, woman or non-binary person (a person who is not exclusively a man or a woman). Given that the non-binary population is small, data aggregation to a two-category gender variable was used to present estimates in this technical report. Individuals in the category “non-binary persons” were distributed into the other two gender categories and are denoted by the “+” symbol. Throughout the report, “men+” designates men (and boys), as well as some non-binary persons, while “women+” includes women (and girls), as well as some non-binary persons.

Population net undercoverage was generally higher for men+ and highest for young adults. In the general population, net undercoverage was highest in the 20-to-34 age group for both men+ and women+. It stood at 9.40% among persons aged 20 to 24, and 7.83% among those aged 25 to 34. For men+, the highest net undercoverage rates were 9.88% among men aged 25 to 34 and 9.44% among men aged 20 to 24. For women+, these rates reached 9.35% among those aged 20 to 24 and 5.67% among those aged 25 to 34. For women+ aged 65 and older and men+ aged 75 and older, the net undercoverage rate was negative, indicating more excess enumerations than persons who were not enumerated.

The net undercoverage rate of the population aged 15 and older was high for separated persons (11.38%), especially for men+ (17.01%). It was also high for single persons (6.76%), i.e., persons who had never legally married and who were not in a common-law union.

The net undercoverage rate for those whose mother tongue is English only was higher than for those whose mother tongue is French only (2.51%, compared with 0.33%). Excluding persons who declared more than one mother tongue, the net undercoverage rate for allophones, persons whose mother tongue is neither English nor French, was higher (6.23%).

Net population undercoverage was slightly more frequent outside CMAs than within them. Nationally, net undercoverage was 3.23% for those who should have been enumerated outside a CMA. This was slightly higher than the net undercoverage rate for those living in a CMA (2.91%).

1.3 Undercoverage

Undercoverage generally refers to persons who were not included as usual residents in the questionnaire that was completed for their usual residence, or persons for whom no questionnaire was completed for their usual residence. For example, persons who regarded their residence as temporary may not have been included as usual residents elsewhere. Persons without a usual place of residence (e.g., the homeless) were also much more subject to undercoverage.

This section presents undercoverage estimates based on the following geographic and demographic variables:

• province or territory of current residence at the time of the census
• age and gender
• marital status and gender
• mother tongue
• CMA of the usual place of residence on Census Day.

Table 1.3 provides the estimated undercoverage regarding the number of persons not included, the standard error of the estimate, the corresponding estimated undercoverage rate and its standard error.

There were some demographic trends in undercoverage. The rate of undercoverage for men+ was one-and-a-half times the rate for women+, or 6.01% compared with 3.95%. The undercoverage rate was highest for young adults aged 18 to 34 (men+ and women+). Among young men+, undercoverage was 12.97% for those aged 20 to 24 and 12.00% for those aged 25 to 34, but it was higher for those aged 35 to 44 than for those aged 18 to 19, contrary to what was typically observed in the past.

In terms of marital status, undercoverage was highest among those aged 15 or older who were separated and not in a common‑law union, at 12.77%. It was also high for single persons not in a common-law union, at 9.48%. In both cases, these rates were higher for men+ than for women+.

Table 1.3 shows that the persons most likely to have been missed in the 2021 Census were men+ aged 20 to 34 who were single (who were never married and not in a common-law union), and separated persons. Mother tongue was also significant for undercoverage rates because for persons who declared only one mother tongue, undercoverage rates were lower among those whose first language was French (2.48%), followed by those whose first language was English (4.48%). For persons whose only mother tongue was neither English nor French, the rate was even higher, at 8.10%.

1.4 Overcoverage

Population overcoverage is the number of excess enumerations in the census counts for persons enumerated more than once (usually twice). This error produces bias because these persons should have been enumerated only once. Examples of overcoverage include children whose parents live separately who were included on each parent’s census questionnaire, persons who do not live with their family for work reasons who are listed on their family’s form and also on the form for the dwelling they live in while working, and students away at school who are listed both by their roommates and by their parents.

This section presents estimates of overcoverage based on the following geographic and demographic variables:

• province or territory of current residence at the time of the census
• age and gender
• marital status and gender
• mother tongue
• CMA of the usual place of residence on Census Day.

Table 1.3 also contains estimates of the number and rate of excess enumerations and the associated standard error.

The standard errors for overcoverage estimates found in Table 1.3 were lower than those for undercoverage. As indicated below, certain demographic trends emerged from overcoverage estimates.

Across the provinces and territories, the estimated overcoverage rate varied less than the estimated undercoverage rate. In fact, the gap between the lowest rate and the highest rate among all provinces and territories was 0.73 percentage points for overcoverage (with the rate varying from 1.68% to 2.41%), whereas this gap was 7.27 percentage points for undercoverage (with the rate varying from 2.64% to 9.91%). Men+ and women+ had similar overcoverage rates, at 2.01% compared with 1.95%. Overcoverage was higher for children and young adults aged 5 to 24, with rates varying from 3.21% to 3.71% for these age groups. As mentioned, these high rates were largely attributable to multiple enumerations of children in shared custody and students.

For marital status, the overcoverage rate was higher for persons who had never been married and who were not in a common‑law union (2.72%). This phenomenon was observed equally among women+ and men+, and it was consistent with the results by age.

In summary, Table 1.3 presents the profile of persons most likely to be counted more than once; they were equally likely to be men+ or women+, and more likely to be a child or a young adult. For persons aged 15 and older, single persons were more likely to be counted more than once.