3.2 Coverage steps

The goal of Steps A, B and C is to establish the exact count of the population on the census reference day.  Two concepts are essential for population coverage: the target population (that is, who to include) and where each person in the target population should be enumerated.  Ideally, each person in the target population should be enumerated only once and at the address where he or she usually lives on Census Day.

A number of changes were made to Steps A and B for the 2008 Census Test to improve respondent understanding of these two concepts.

Top of Page

3.2.1 Change order of pages 2 and 3

The placement of questions on pages 2 and 3 was reversed in the alternate version of the 2B questionnaire compared to the control version.  The reason for this switch was to reduce the number of people who incorrectly identify themselves as temporary or foreign residents (i.e., false positive responses), and to improve the process of confirming each reported dwelling that is occupied solely by temporary residents or foreign residents (also called TRFR), thus improving data quality.  Refer to Step B for the results of the TRFR analysis.

Top of Page

3.2.2 Step B – Question 1, Number of persons living in the dwelling

Figure 3.1 Step B, Question 1

Source: Statistics Canada, 2008 Census Test, 2B.1 questionnaire.

A change in the alternate version of the questionnaire was introduced based on feedback received during qualitative testing.  We learned that some participants did not include themselves among the count of household members because they thought they had already provided their name in Step A.  They would complete the questionnaire for other members of the household and if there were no others, would simply explain: "There is no one else, I live alone."  To address this situation, the reminder 'including yourself' was moved from the beginning of the question to the end, as shown in Figure 3.2 below.

Figure 3.2 Step B, Question 1

Source: Statistics Canada, 2008 Census Test, 2B.2 questionnaire.

Response to this question is important because it is used for editing during the processing of the questionnaires.  The edit ensures that the number of people reported on the questionnaire are consistent (i.e., the number of people in Step B1, the number of names listed in Step B2, and the persons enumerated in Step E, Question 1 are the same).  These three fields must correspond otherwise the questionnaire fails the coverage edit and must be manually examined.

Despite the introduced change, the question did not perform as well as expected during the 2008 Census Test and a thorough examination was undertaken to determine the reasons for the high level of non-response to the question.

Tables 3.1 and 3.2 show the non-response rates for Step B, Question 1.  The rate is calculated only for households that returned a paper questionnaire.  The question is mandatory in the Internet questionnaire because it is used to create a record for each person, much like the column for each person that appears on the paper questionnaire.

Table 3.1
Non-response rate for Step B, Question 1 – Paper returns (unweighted)

Table 3.2
Non-response rate for Step B, Question 1 – Paper returns (weighted)

The item non-response rate was high in all of the self-enumerated panels.  Subsequent qualitative testing showed that the majority of participants did not answer this question simply because they did not see the response boxes and moved right into listing the names of household members.  Interviewers that participated in the questionnaire follow-up for the 3B panels also missed this item.  As a result, the format of the question was modified for the 2009 Census Test as shown in Figure 3.3.

Figure 3.3 Step B, Question 1

Source: Statistics Canada, 2009 Census Test, 2B questionnaire.

The change in format proved to be very effective.  The non-response rates for the general population (the 2A and 2B panels in the 2009 Census Test compared to the 2A2 and 2B2 panels in 2008 Census Test) decreased to 7.3% from 27.6% and to 12.5% from 29.4%, respectively.  As a result, the placement of the response boxes below the question instead of beside the question text was recommended for the 2011 Census.  A very small change was made to the question itself.  The word 'examples:' within brackets was removed from the instructions that follow the question.

Top of Page

3.2.3 Step B – Question 2, List of household members

Two changes were made to Step B, Question 2.  The first was a change in the wording of the instruction.  The second was to separate the text boxes to identify clearly which text boxes were for writing in the family name and which were for writing in the given name.

(a) Change in wording of instruction

In the 2006 Census and the control questionnaires for 2008, the following text appeared as the second sentence of the instruction:

'Begin the list with an adult followed, if applicable, by that person's spouse or common-law partner and by their children who usually live here.'

In the 2008 test (alternate questionnaire), the first part of this sentence, 'Begin the list with an adult' was replaced with 'When applicable, begin with a parent or spouse.'  Secondly, a sub-text was placed below the headings down the left side of the page for Person 1 and Person 2, to reinforce this instruction.  The relevant section of the page is shown in Figure 3.4, below.

Figure 3.4 Step B, Question 2

Source: Statistics Canada, 2008 Census Test, 2B.2 questionnaire.

The reason for the change in emphasis, from starting with any adult to starting with a parent or spouse, was related to the changes made in the question on Relationship to Person 1 to better identify stepchildren.  Refer to Section 3.6.3 for more information on the changes to that question.  It was considered important to encourage respondents to list household members in a way that the parent or parents—including a potential step-parent—would be placed in the positions of Person 1 and Person 2, followed by those children.  This would avoid having an adult who is neither the parent nor step-parent of the children in the position of Person 1 or Person 2, which could cause a misreporting of the parent-child relationships.

The above suggested order remains the preferred order of listing in the 2011 Census to obtain accurate reporting for all three new son-daughter categories for Relationship to Person 1.  Some concerns about the new wording were raised during qualitative testing.  Relatively few participants in qualitative testing read the instruction.  Furthermore, among some young participants who did read the instruction, it was observed that they wondered, at least briefly, if they should report their parent who lived in a separate household—which was not the intended meaning.

With these concerns, it was decided to change back to the wording used in the 2006 Census that instructs respondents to begin the list with 'an adult.'  The question was modified for the 2009 Census Test and remains the same on the 2011 Census (Figure 3.5).

Figure 3.5 Step B, Question 2

Source: Statistics Canada, 2009 Census Test, 2B questionnaire

(b) Separation of text box spaces for family name and given name

The need to improve the quality of names collected on the paper questionnaires prompted a change to the structure of the text boxes.  Improvement to the names will facilitate census follow-up and evaluation activities, as well as sample selection for specific populations.

In the 2008 Census Test, the text boxes were divided into two sets, 15 boxes for family name and 12 for first given name as seen in Figure 3.4.  Data capture operators manually captured names that exceeded the number of available boxes or contained more than one letter in a box, if detected, up to a maximum of 80 characters, thus minimizing the loss of full names.

Although separating the family and given names required more manual data capture, the recommendation was made to keep this separation for the 2011 Census as it is easier to distinguish and capture names, thus leading to better quality.

Top of Page

3.2.4 Step B – Dwelling occupied by temporary residents and foreign residents

Changes were made to the instructions of the alternate 2A.2 and 2B.2 questionnaires to help respondents self-identify themselves into three separate groups: (1) temporary residents^{Footnote 1} (2) foreign visitors, and (3) government representatives of a foreign country.  We refer to these temporary and foreign residents as 'TRFR.'

The title of the section was also changed to refer to the dwelling address.  This change emphasized that every person living in the dwelling must be a temporary resident, foreign visitor, or a government representative of another country.  (If the dwelling is the usual residence of at least one person, then the person(s) must to be enumerated.) A bulleted list was created thus allowing respondents to specify the appropriate response category that described the dwelling and its occupants.  Finally, an address field was added to improve and possibly reduce the amount of TRFR follow-up activities.  These changes can be visualized in Figure 3.6 (control question) and Figure 3.7 (alternate question).

The extent of analysis for dwellings occupied by TRFR was limited in the 2008 Census Test for two reasons: (1) the dwellings known to be occupied TRFR during the 2006 Census were excluded from the target population, and (2) no follow-up action was taken to verify the status of the households that identified themselves to be TRFR.

Figure 3.6 Step B, item 3 instructions for identifying temporary residents and foreign residents (TRFR), control questionnaire

Source: Statistics Canada, 2008 Census Test, 2B.1 questionnaire.

Figure 3.7 Instructions for identifying temporary residents and foreign residents (TRFR), alternate questionnaire

Source: Statistics Canada, 2008 Census Test, 2B.2 questionnaire.

Proportionally more respondents entered information in the TRFR section of the control questionnaire (Figure 3.6), where the question was located on page 2 (even-numbered page appearing on the left side), compared to the number of respondents that completed the alternate questionnaire (Figure 3.7), where the question was located on page 3 (odd-numbered page appearing on the right side).

Subsequent qualitative testing confirmed that the first thing most participants did when they opened the questionnaire was to look for a place to answer the first question, not realizing that the question did not apply to them.

Although the content test could not be used to draw conclusions about the TRFR question, the decision was made to include the alternate question in the 2009 Census Test.

2009 Census Test

The sample for the 2009 Census Test did not exclude temporary and foreign residents from being selected.  The test also implemented non-response follow-up (NRFU) for dwellings that did not return a questionnaire, and conducted failed edit follow-up (FEFU) procedures for questionnaires that failed edit steps during initial processing of questionnaires.  As a result, the 2009 test was better source for evaluating the TRFR question.

In total, 1,075 respondents provided information that identified their dwelling to be occupied by TRFR.  The majority of respondents (75%) completed a paper questionnaire, 15% completed the Internet questionnaire, and the remaining 10% responded during non-response follow-up (NRFU) or during calls from the Census Help Line.

A total of 955 potential TRFR cases were sent to FEFU for resolution.

• Interviewers completed interviews with 68% of the households during the follow-up period.
• Overall, 42% were confirmed to be true TRFR households.  (The confirmation rate was 79% for respondents that completed an online questionnaire, compared to 33% for those who completed a paper questionnaire.)
• Among cases where all of the required TRFR fields were completed and no usual residents were enumerated in the questionnaire, 86% of cases were confirmed to be true TRFR households.
• Among cases where the TRFR fields were completed and there were usual residents present in the questionnaire, only 15% were confirmed to be true TRFR households.

Questionnaires that had no evidence of a usual resident (e.g., no persons listed anywhere in the questionnaire or the TRFR address differed from the address in Step A), were confirmed to be true TRFR cases 89% of the time.  This key finding suggested that we could potentially exclude these 'clear-cut' cases for follow-up in the 2011 Census.

For the 2011 Census, a recommendation was made to: change the order of pages 2 and 3, modify the TRFR instructions, and modify the rules for sending cases to FEFU that would take into account the above results.