Coverage Technical Report, Census of Population, 2016
The 2016 Census required the participation of the entire population of Canada, roughly 36 million people over a territory of nine million square kilometres. Although data collection and processing have to meet rigorous quality standards, it is impossible to eliminate all errors. To use census data correctly and appropriately, it is important to understand the conceptual framework and the definitions used to conduct the census, and the data collection and processing methods. Users also need to know the main sources of error and, where possible, the size of the errors and any unusual circumstances that might limit the usefulness or interpretation of census data. With this information, users can assess the risks associated with using census data to draw conclusions or make decisions.
This technical report examines coverage errors in the 2016 Census. There are two types of coverage errors. The first, called population undercoverage, refers to excluding persons who should have been enumerated. The second, called population overcoverage, refers to either including persons who should not have been enumerated or enumerating persons more than once. In the first instance of overcoverage, the overall error is considered negligible. Undercoverage is generally more common than overcoverage. The net impact of undercoverage and overcoverage on the size of a population of interest is population net undercoverage. Net undercoverage is the number of persons excluded who should have been enumerated (undercoverage) less the number of extra enumerations of persons who were enumerated more than once (overcoverage). Coverage errors are among the most significant types of errors, since they affect the accuracy not only of the counts for the various census universes, but also of all the census data that describe the characteristics of these universes.
Census coverage errors are measured using three studies. The first, the Dwelling Classification Survey (DCS), examines coverage errors resulting from errors in classifying dwellings. Census data are adjusted for this type of coverage error. The second, the Reverse Record Check (RRC), measures population undercoverage, while the third, the Census Overcoverage Study (COS), measures population overcoverage. Census data are not adjusted for the coverage errors measured by the RRC and the COS. Rather, Statistics Canada uses net undercoverage estimates to produce demographic estimates. The 2016 Census studies are quite similar to the 2011 studies, with some changes and improvements.
Census data users should be aware that the presence of coverage errors in the 2016 Census means that census products are subject to incomplete enumeration or double-counting. Undercoverage, for example, is higher for young adult males. For 2016 Census coverage error estimates for various demographic and geographic levels and groupings, see Section 1.
Section 2 covers the 2016 Census conceptual framework and provides definitions of the population universe, the dwelling universe and the usual place of residence that the census aims to measure. Section 3 describes coverage errors, their cause, census practices that minimize them and the conceptual framework used to evaluate them. It also introduces census coverage studies. Section 4 and Section 5 describe the data collection and processing methodology used in the 2016 Census.
Section 6 presents the methodology used during the 2016 DCS and the results obtained. Coverage error rate estimates are produced only for the population universe. Section 7 and Section 8 describe the methodology used for and the results of the 2016 RRC and 2016 COS, respectively. Section 9 shows how the results of the RRC and the COS were combined with census data to produce population coverage error estimates and the associated standard errors.
Section 10 covers the results of evaluations performed for the RRC and the COS, and the evaluation of the error of closure. “Error of closure” refers to the difference between demographic estimates whose base population is the 2011 Census counts, adjusted for net undercoverage, and 2016 Census data, also adjusted for net undercoverage.
Statistics Canada has conducted census population coverage studies since the first RRC, which dates back to the 1961 Census.Note 1 Section 11 provides a chronological review of coverage errors from the 1971 Census to the 2016 Census.
Section 12 covers additional topics, including the concept of persons not enumerated, and participation by Indian reserves and Indian settlements in the census.
This report was prepared by Melanie Abeysundera, Karen Bruce, Lucia Debroy, Heather Farr, Michel Parenteau, Martin Provost and Martin St‑Pierre of the Statistical Integration Methods Division, and Julien Bérard‑Chagnon of the Demography Division.
Many people made valuable comments on preliminary versions, which improved the content and readability of the final report.
For additional information on census concepts, variables and geography, please see the Dictionary, Census of Population, 2016, Catalogue no. 98-301-X. For additional information about the census process, please see the reference materials, 2016 Census.