Aboriginal Children's Survey, 2006: Concepts and Methods Guide
Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.
3.0 Survey objectives
4.0 Survey development
5.0 Survey design
6.0 Data collection
7.0 Data processing
8.0 Data quality
10.0 The relationship between the ACS and the Census
This guide is intended to provide an understanding of the concepts and methods used in the 2006 Aboriginal Children's Survey (ACS), which was conducted from October 2006 to March 2007.
The Aboriginal Children's Survey (ACS) provides data on the development and well-being of Aboriginal children under 6 years of age (First Nations children living off reserve, Métis children and Inuit children) living in urban, rural and northern locations throughout Canada.
Aboriginal children living in Indian settlements and reserves in the 10 provinces were not included in the 2006 ACS data collection. In the three territories, all First Nations children were included in the target population. Some First Nations1 communities in Quebec2 were also included.
Technical details on sampling, processing and data quality are included in this guide. Further, the guide explains the relationship between the ACS and the 2006 Census and cautions users as to important differences in the data produced from these two sources.
Appendix 1 contains a glossary of terms that relate to the ACS. Answers to some frequently asked questions are provided in Appendix 2. Links to the 2006 ACS questionnaires are found in Appendix 3.
Survey of Northern Children
The Survey of Northern Children (SNC) was originally a component of the National Longitudinal Survey of Children and Youth. In 2006, children selected for the Survey of Northern Children were surveyed using the Aboriginal Children's Survey / Survey of Northern Children 2006 (Children - aged 0 to 5) questionnaire (see Appendix 3).
The SNC targeted all children (without distinction between Aboriginal and non-Aboriginal children) under 6 years of age living in Nunavut, Northwest Territories and Yukon.
The Aboriginal Children's Survey (ACS) is a post-censal survey conducted for the first time in 2006. The data collected allows Aboriginal organizations, community planners, service providers, parents, governments and researchers to:
- Honour and acknowledge Aboriginal children (needs, values and cultural heritage)
- Inform decision-making (program and policy planning and development)
- Support academic research (educators and researchers)
The survey was developed by Statistics Canada in partnership with Human Resources and Social Development Canada and Aboriginal advisors from across the country. A unique process was used involving direct participation of parents, front-line workers, early childhood educators, researchers and Aboriginal organizations.
Technical Advisory Group (TAG), which consists of educators, researchers and other professionals in the field of Aboriginal research and early childhood development (ECD), provided guidance for the ACS on an ongoing basis.
The target population for the ACS was previously covered by the Aboriginal Peoples Survey (APS), Children and Youth Component in 2001 and 1991, which collected data for Aboriginal children 0 to 14 years of age.
3.0 Survey objectives
There is currently little data available about the health and development of Aboriginal children under 6 years of age. The Aboriginal Children's Survey (ACS) was designed to address this data gap.
Based on recommendations from the TAG, the ACS 2006 is holistic in nature and collected information on a wide range of topics, including child's health, sleep, nutrition, motor, social and cognitive development, nurturing, child care, school, language, behaviour, and activities. Since the child's environment is important to their development and well-being, some information was collected on the child's parent(s) or guardian(s) and their neighbourhood or community.
4.0 Survey development
The Aboriginal Children's Survey (ACS) was developed with the advice and input of Aboriginal people, organizations, governments and others. Discussions were held with major stakeholders to ensure that the ACS is relevant to the needs of a variety of different data users.
The development process for the ACS began with discussions with national Aboriginal organizations followed by regional content discussions, both of which provided an understanding of the broad areas of interest and data needs. A Technical Advisory Group (TAG) was established to obtain the advice, guidance and perspectives of Aboriginal peoples throughout the development and implementation of the ACS. As well, federal, provincial, and territorial governments were involved in identifying priority areas to be included in the questionnaire.
4.1.1 National Aboriginal Organizations (NAOs)
Statistics Canada works closely with National Aboriginal Organizations on all of its Aboriginal data projects. The National Aboriginal Organizations who have provided input into the ACS are:
- Congress of Aboriginal Peoples (CAP)
- Inuit Tapiriit Kanatami (ITK)
- Métis National Council (MNC)
- National Association of Friendship Centres (NAFC)
- Native Women's Association of Canada (NWAC)
At the beginning of the survey process, Statistics Canada met with representatives from the National Aboriginal Organizations in order to assess the feasibility of undertaking a national survey about Aboriginal children. These organizations provided input on their data needs and have provided advice throughout the process.
4.1.2 Regional content discussions
In 2004, Statistics Canada held a number of one-day meetings with representatives from Aboriginal organizations, service-providing agencies, researchers and parents of Aboriginal children in selected regions across Canada to discover important issues in the early development of Aboriginal children, the appropriateness of possible topic areas, and regional issues to be considered.
4.1.3 Technical advisory group
A Technical Advisory Group (TAG) was established to provide advice and guidance throughout the survey process. This group was composed of a cross-section of Aboriginal and non-Aboriginal educators, researchers and other professionals in early childhood development.
In collaboration with the TAG, Statistics Canada developed a survey specifically designed for young Aboriginal children including questions that provide a cultural context. The TAG was invaluable in recommending content and processes that respect First Nations, Métis, and Inuit values.
4.1.4 Federal / Provincial / Territorial Governments
A Steering Committee comprising representatives from Human Resources and Social Development Canada (HRSDC), Health Canada (HC), and Indian and Northern Affairs Canada (INAC) was established to provide ongoing strategic direction to the ACS.
Prior to developing the survey content, Statistics Canada contacted federal, provincial and territorial governments to determine the data needs and priorities of these levels of government. Governments were asked to identify areas or topics of importance to their department or region.
4.2 Survey content
4.2.1 Aboriginal Children's Survey questionnaire
The Aboriginal Children's Survey (ACS) is holistic in nature and covers a wide variety of topics. Cultural elements are woven throughout the various sections of the questionnaire.
The questionnaire was administered to a parent or guardian of children under 6 years old. The following is a list of the sections and some of the key variables. See Appendix 3 for links to the 2006 ACS questionnaire.
- Aboriginal Ancestry
- Aboriginal Identity
- Marital status
- Relationship to child
- General health
- Birth weight
- Contact with health professionals (paediatrician, public health nurse, and other)
- Chronic conditions
- Incidence of injuries
- Types of foods child eats
- Sleeping Habits
- Expressing needs
- Expressing needs
- Telling stories
- People involved in raising the child
- School attendance
- School grade
- Type of childcare
- Reasons receiving child care
- Reasons not receiving regular child care
- Language of childcare
- Cost of childcare
- Ability to understand and speak languages
- Home language
- Exposure to Aboriginal language
- Conduct problems
- Emotional symptoms
- Peer problems
- Pro-social behaviour
- Play activities
- People with whom the child plays
- Understanding First Nations, Métis or Inuit culture
- Aboriginal ancestry and identity
- Main Activity
- Mother tongue
- General health rating
4.2.2 Strengths and difficulties questionnaire
The Aboriginal Children's Survey (ACS) includes the Strengths and Difficulties Questionnaire (SDQ), developed by Robert Goodman (2000). The SDQ is made up of 25 items (included as questions K01A to K01Y on the ACS). These 25 items are designed to be grouped into 5 sub-scales that measure: conduct problems, inattention-hyperactivity, emotional symptoms, peer problems, and pro-social behaviour.
The Aboriginal Children's Survey (ACS) file includes the 25 items of the SDQ; however the 5 sub-scale groupings have not been included on the analytical file. Please note that: 1) the version of the SDQ included on the ACS was designed for children aged 3 to 4 years, although it was applied to 2 to 5 year olds in the ACS; 2) the SDQ was developed to be self-administered by parents, while the ACS was conducted in a personal-interview format; and 3) the SDQ was not specifically designed to assess Aboriginal children in Canada. Although the validity and reliability of the SDQ has been widely confirmed internationally, further research is required to explore the applicability of the SDQ and the 5 sub-scales to Aboriginal children in Canada (including each of the three Aboriginal groups of First Nations living off reserve, Métis and Inuit).
4.2.3 Census topics appended
Some information that was obtained from the 2006 Census has been appended to the ACS analytical file to provide a very rich and detailed data set for analysis.
The following Census variables have been appended to the ACS analytical file.
- Period of construction of dwelling
- Structural type of dwelling
- Is anyone in the household a farm operator?
- Gross rent
- Marginal dwelling indicator
- Number of household maintainers
- Owner's major payments
- Registered condominium status
- Number of rooms
- Is dwelling in need of repair?
- Tenure of dwelling
- Value of dwelling
- Census family total income
- Economic family total income
- Household total income
- Low income status (before taxes)
- Census subdivision type one year ago
- Mobility status - place of residence one year ago
- Census subdivision of residence one year ago
- Rural-urban place of residence one year ago
Note: Because these variables were obtained from the 2006 Census responses for the children surveyed on the ACS, they refer to the child's situation on the day of the Census, that is, 16 May 2006. Users should be aware that in some cases, the child could have moved or the composition of the household could have changed between Census day and the date of the ACS interview, so that some of the information provided by the census data may not always be reflective of the child's situation when the ACS interview took place.
5.0 Survey design
5.1 Target population
5.1.1 Identifying Aboriginal Children
The target population for the ACS includes all children in Canada with North American Indian, Métis or Inuit identity or ancestry, under the age of 6 years as of October 31, 2006, excluding children living in Indian settlements or on-reserve. Children living in institutions were not included. Although children living on-reserve were not included in the provinces, all Aboriginal children living in the territories and children in some First Nations communities in Quebec were included.
The target population for the Survey of Northern Children includes all children (Aboriginal and non-Aboriginal) living in the territories, under the age of 6 years as of October 31, 2006. Children living in institutions were not included.
The Aboriginal Children's Survey (ACS) is a postcensal survey, which means that the ACS sample was selected based on reported answers to the Census questionnaire. More precisely, the ACS sample was selected based on reported answers to four screening questions on the Census long form, which has two main versions, the 2B form and the 2D form.
The 2B form is completed by self-enumeration and is administered to approximately one in five households in most parts of Canada (2B regions). Other than the basic demographic questions, the 2B form asks questions on labour activity, income, education, activity limitations, citizenship, housing, ethnic origin and so on.
The 2D form, identical in content to the 2B form except for some adaptation of examples, is administered by personal interview to all households in remote areas and Indian reserves (2D regions).
Other households in Canada receive the short form, which only contains basic demographic questions (name, sex, date of birth, legal marital status, common-law status, relationship to person 1, first language learned in childhood and consent question to make data public in 92 years).
The four screening questions used to identify the Aboriginal population are: the ethnic origin question (question 17); the Aboriginal self-reporting question (question 18); the Indian band / First Nation membership question (question 20); and the Treaty or Registered Indian question (question 21).
The derived Aboriginal identity concept refers to those persons who reported identifying with at least one Aboriginal group, that is, North American Indian, Métis or Inuit, and / or those who reported being a Treaty Indian or a Registered Indian, as defined by the Indian Act of Canada, and / or those who reported they were members of an Indian band or First Nation. The Aboriginal identity population is derived from 3 questions (Questions 18, 20 and 21).
The addition of non-Aboriginal children in the territories allowed for integration with the Survey of Northern Children (SNC) which targets estimates for all children under 6 without distinction between Aboriginal and non-Aboriginal children.
5.1.2 Survey reference date
October 31, 2006 was selected as the reference date for the ACS. This date approximately corresponds to the beginning of data collection for the survey (October 23 was the official start date). The age is determined as of this reference date and is used in the skip patterns of the questionnaire.
5.1.3 Census frame
A sampling frame provides a means of accessing the population to be covered by a survey. The Aboriginal Children's Survey (ACS) frame was built in a series of steps. The frame was constructed for both the ACS and the APS. In the first step, a list was created containing all individuals falling in the target population according to answers reported to the four screening questions of the Census long form (children and adults). (Also included in this step were the non-Aboriginal children under 6 living in one of the territories.)
In a second step, household members of the Aboriginal children selected in the first step were added to the list. These correspond to additional household members belonging to an Aboriginal household (household containing at least one Aboriginal person). These additional individuals could be potentially added to the target population if they had missing information to the screening questions (at the time of sample selection, Census data were not imputed and could therefore be missing). These individuals were also used as potential contact persons to trace the selected child(ren) in the household.
In a third step, missing information on the screening questions was imputed. Missing information on screening questions (or "filters") belonging to an Aboriginal household have a good chance of being Aboriginal people as well. On the other hand, individuals with missing filters not belonging to an Aboriginal household have a very small chance of being Aboriginal people and were not considered as potential additions to the frame. In general, the imputation rules looked at the household composition for individuals with non-missing filters. Individuals with missing filters were imputed as being Aboriginal people if at least 50% of the household members with non-missing filters were Aboriginal people.
In the fourth and the final step, all Aboriginal children under 6 (based on the imputed data), were kept on the frame for ACS as well as all non-Aboriginal children in the territories. Aboriginal people with missing age responses on the Census form were all excluded from the ACS and were included on the APS frame (more likely to be older than 5 than being under 6)
5.2 Sampling design
5.2.1 Domains of estimation
Domains of estimation are groups of units for which estimates are targeted. The domains of estimation for the ACS were selected to provide the maximum level of detail possible to users given the very small size of the target population. Ideally, estimates would be produced by province / territory / region, age group (0-1, 2-3 and 4-5) and Aboriginal group (First Nations, Métis and Inuit).
The Aboriginal identity groups were defined as follows:
- North American Indian (NAI) only - individuals reporting only NAI to question 18
- Métis only - individuals reporting only Métis to question 18
- Inuit only - individuals reporting only Inuit to question 18
- Multiple identity – individuals reporting more than one group to question 18
- Registered Indian or band member only – individuals with a positive answer to question 20 or question 21 but 'No' to question 18
Some of the combinations of province / territory / region, age group and Aboriginal group would include a very small number of children, which could cause either confidentiality problems or poor-quality estimates (or both). The domains of estimation were designed according to estimated population counts of the identity population as well as the estimated number of long-form records available to produce the estimates. This was based entirely on the counts from the 2001 Census.
Estimates for the Inuit (Inuit only) were required for each of the four Inuit Land Claim regions, namely Nunatsiavut, Nunavik, Inuvialuit and Nunavut. The regions of Inuvialuit and Nunatsiavut are considerably smaller than Nunavik and Nunavut. Therefore, for these two regions, age groups were combined. As a general rule, it was decided, for confidentiality reasons, to target estimates for populations of at least 200 children. When this could not be met, age groups were combined.
For the rest of the population, estimates would be produced for each age group for First Nations and Métis separately. Due to the small numbers, this would not be possible. In terms of geography, the Atlantic provinces had to be grouped together. In terms of Aboriginal groups, only Ontario and the four Western provinces contained a large enough number of Métis children to consider separate estimates for the Métis. In the Atlantic region, Quebec, Northwest Territories and Yukon, all Aboriginal groups were combined. In addition, in Yukon, following the rule of a minimum population of 200 children per age group (the 0-1 age group being the smallest), all age groups had to be combined.
Estimates are to be produced within each domain of estimation. In general, within each domain of estimation, the target was to estimate a characteristic present for no less than 10% of the population with a coefficient of variation (CV) of 20%. The CV is a measure of precision of the estimate which is described in Section 8.1 (Sampling Errors). The ability to achieve this CV would depend on factors such as the population size, the number of long forms available, the expected response rate, the expected number of false positives ( children who were reported to be Aboriginal on the Census but who were reported to be non-Aboriginal on to ACS), the expected sample loss due to the constraint of selecting no more than two children per household, the expected loss due to the reduction of overlap with other postcensal surveys and the elimination of the overlap with the National Longitudinal Survey of Children and Youth (NLSCY) (section 5.3). In certain domains, it was not possible to achieve this precision. In these cases, a CV of 25% was targeted.
Table 1 gives the targeted CV and targeted / expected response rate (R) for each domain of estimation. In this table, the domain "Atlantic Other" refers to the Atlantic provinces excluding Nunatsiavut. The domain "Quebec Other" refers to Quebec excluding Nunavik. Finally, the domain "Northwest Territories Other" refers to the Northwest Territories excluding Inuvialuit.
Targeted coefficient of variation and response rate by domain of estimation for the identity population, 2006
5.2.2 Sampling plan
As mentioned, the ACS selects its sample from the Census long form sample (either from the 2B or the 2D version of the long form). Outside Indian reserves, the 2D covers the Northern part of each province and the three territories with the exception of Yellowknife and Whitehorse which use the 2B form. In 2D regions, all households receive the 2D version of the long form. In 2B regions (all parts of Canada outside 2D regions), a systematic sample of approximately one in five households receives the 2B version of the long form within each Collection Unit (CU).
Once the frame has been constructed, it is then stratified according to the domains of estimation and is further stratified by 2B / 2D regions. A simple random sample is then selected within each domain of estimation crossed by 2B / 2D regions. Since the ACS sample is a sample of the long form sample, its sample design is called a two-phase sample, where a sample of households is selected in the first phase and a subsample of children is selected in the second phase.
5.3 Overlap with other surveys
In order to control respondent burden, it was decided to reduce the overlap between the ACS and the other postcensal surveys as well as the National Longitudinal Survey of Children and Youth (NLSCY). For the Aboriginal Children's Survey (ACS), a total of 835 units were removed from the sample and treated as a particular form of non-response in the weighting process. These different sample losses had been estimated before selecting the final sample and the original sample size was increased to compensate for this loss.
5.3.1 Overlap with other postcensal surveys
In 2006, five postcensal surveys were conducted at approximately the same time, the ACS, the Aboriginal Peoples Survey (APS), the Participation and Activity Limitation Survey (PALS), the Survey on the Vitality of Official-Language Minorities (SVOLM) and the Maternity Experiences Survey (MES). All of these surveys selected their sample from the Census and most of them only from answers to the Census long form. This means that a given household could potentially have been selected for up to five surveys if the household had members of all target populations. Although very unlikely, a household could have been selected for three or four surveys in some cases. Also, more than one person in each household could have been selected for the same survey. The absence of a procedure to reduce the overlap at the household level could have represented a very high response burden for many households.
Consequently, rules were used to limit the overlap between the different surveys once the samples were selected. The idea was to limit the number of surveys to two per household and to three interviews per household. In certain cases, four interviews per household (two for each of two surveys) were allowed. In a first step, the number of surveys per household was reduced to a maximum of two. If a household was initially assigned to more than two surveys, two surveys were selected at random. In a second step, the number of interviews per household was limited to three or four using another random procedure.
5.3.2 Overlap with National Longitudinal Survey of Children and Youth
The National Longitudinal Survey of Children and Youth (NLSCY) content is closely related to that of the ACS. For this reason, it was decided to exclude some selected ACS children falling in households of certain NLSCY cohorts. There were also certain children selected in the 0-1 NLSCY cohort overlapping with the ACS sample who were dropped from the NLSCY sample.
5.4 Sample sizes
The initial sample size excluding the First Nations communities in Quebec was 18,307 children and after the overlap reduction with other surveys, this number was 17,472 children. The distribution of these 17,472 children in the various domains of estimation is given in Table 2 by geographical domain and type of population (identity, ancestry-only and non-Aboriginal) with their corresponding observed response rates. It should be noted that the number of children also include those whose parent or guardian consented to participate in the survey but who were reported as being non-Aboriginal in the ACS (false positives).
6.0 Data collection
The Aboriginal Children's Survey was conducted between October 2006 and March 2007.
6.1 Mode of collection
The Aboriginal Children's Survey was collected using a paper questionnaire and the respondent was the parent or guardian of the selected child.
The survey was conducted using personal interviews in Inuit regions, Labrador and in the Northwest Territories (except Yellowknife). Telephone interviews were conducted elsewhere in Canada. In a number of locations, personal interviews were undertaken when people could not be reached by telephone.
6.2 Aboriginal languages
The ACS questions were translated into 5 Inuktitut dialects (Innuinaqtun, Inuvialuktun, Labrador, Nunavik and Nunavut) and 2 dialects of Cree. Interpreters were hired when required. Translators were hired when requests were received for the survey to be conducted in other Aboriginal languages.
6.3 Coordination with the Aboriginal Peoples Survey
The collection of the Aboriginal Peoples Survey (APS) occurred during the same period of time as the ACS. In order to keep respondent burden to a minimum, the collection of these two surveys was coordinated. For households who were selected for both surveys, the surveys were conducted by the same interviewer during the same telephone contact or personal visit, whenever possible.
7.0 Data Processing
7.1 Data capture
Data capture was carried out at the head office in Ottawa. Two methods, optical character recognition (scanning) and key entry, were used to capture the questionnaires. Checkboxes and numeric write-in responses (e.g. date of birth) were captured by scanning, while other write-in responses were captured by key entry. Questionnaires were recaptured when data quality fell below acceptable standards. As well, some abnormalities created by the optical reading system were identified and corrected during editing.
The first stage of error detection was done during the data collection. Interviewers were asked to check their questionnaires page by page ensuring that everything had been filled in correctly and clearly and to ensure that skips had been followed correctly. In cases where questions were incorrectly missed, interviewers were instructed to contact the respondent again to obtain the missing information.
The second stage of survey processing involved editing all the survey records according to pre-specified edit rules to check for errors, gaps and inconsistencies in the survey data. Validity checks on each variable were made to ensure, for example, that numerical answers to certain questions fell within acceptable logical ranges and that invalid multiple responses to certain questions were identified. Checks were also made to ensure that the questionnaire flows were followed properly and that portions of the questionnaire that were to be skipped in the interview because of a previous answer were in fact skipped. Inconsistencies between related questions were also corrected.
Where errors were found, the erroneous information was replaced by a "not stated" code, or corrected based on the answers to other questions. Although the corrections were generally done in an automated way, analysts reviewed some problematic situations on a case by case basis.
Finally, a macro-level verification was done by analyzing frequency distributions to identify anomalies (for example, missing categories or unusually large frequencies).
7.3 Processing the household roster
The processing of the household roster (section B) of the ACS questionnaire requires special mention. The roster asked, for each person living in a given household, their date of birth, age, sex, marital status, and relationship to the selected child(ren). If more than one selected child lived in the household, the roster was to be completed only once.
During the processing of the data, an unexpectedly high degree of non-response to the roster questions was observed. In particular, the information about who was the respondent for the selected child was often not provided. This was especially a problem for households having two selected children. In many cases, the paper questionnaires were examined in an effort to resolve these cases or to make informed assumptions about the situation. A series of household-level variables (e.g. number of generations in child's family) were derived based on this section and it is mainly these variables, rather than the responses to the individual roster questions, that were retained on the analytical file. The only individual-level data from the roster that were retained on the analytical file are the age of the selected child, and the age, sex, marital status and relationship of the respondent to the child.
In a sample survey, each selected person represents not only himself or herself, but also other persons who were not sampled. Consequently, a weight is associated with each selected person to indicate the number of persons that he or she represents. This weight must be used for all estimations. For example, in a simple random sample of 2% of the population, each person represents 50 persons in the population. The initial weight is then adjusted for such things as non-response and discrepancies between the characteristics of the sample and known totals for the target population (post-stratification adjustment). In fact, six steps were used in the weighting process.
7.4.1 Initial weights
The initial weight was the inverse of the inclusion probability (probability of falling in the sample). The initial weight was the product of two components: the inverse of the stratum sampling fraction and the inverse of the initial Census sampling fraction. The stratum sampling fraction is calculated as the number of children selected in each stratum divided by the number of long forms available in that stratum on the Census frame. The initial Census sampling fraction, which is unique to each Collection Unit (CU), is calculated as the number of completed long forms divided by the total number of short and long forms for that CU (usually slightly smaller than 1/5 in 2B regions and slightly smaller than 1 in 2D regions because of non-response).
7.4.2 Adjustment for overlap with other surveys
As mentioned in Section 5.3, 835 children were lost from the initial sample of size 18,307 due to the reduction of the overlap with other surveys. To compensate for that loss, a simple ratio adjustment was applied by population type (identity, ancestry-only and non- Aboriginal), Aboriginal group and age group. That is, for each of these groupings, the sum of the initial weights was calculated over the full initial sample and over the remaining sample of 17,472 children after deduction of the overlap. Initial weights were then multiplied by these factors for the remaining sample to obtain the adjusted weights. The adjusted weights of the units removed were set to 0. Hence, the sum of the adjusted weights for the remaining units adds up to the sum of the initial weights within each combination.
7.4.3 Adjustment for units selected in Aboriginal Peoples Survey
A small number of individuals selected for the Aboriginal Peoples Survey were in fact "in scope" for ACS and ended up completing the ACS questionnaire. This was due to errors in the Census date of birth. Although interviewers were not supposed to convert an APS to an ACS questionnaire, a small number of such interviews were done. In order to keep these interviews, a special procedure was used to assign them a weight.
Looking at the APS strata in which these individuals were selected, it was possible to assign the corresponding ACS strata in which they would have fallen had the correct Census date of birth for these children been available. These individuals were initially assigned the average weight (weight adjusted for the overlap loss) of the corresponding stratum. The weights of all children in these strata were then slightly proportionally decreased to preserve the stratum totals.
7.4.4 Adjustment for out of scope units
Some individuals were found to be out of scope for reasons other than reporting that the child was not an Aboriginal person. In fact, 162 individuals were not under the age of 6 years, 10 were deceased and one was no longer living in Canada. Individuals who did not fall into the target group for the survey may have been cases with an error in the Census date of birth. The Census age available for these individuals had a tendency to be proportionally higher for the 0-year-olds and 5-years-olds. The deceased individuals were proportionally higher for the 0-year-olds. Since most out of scope units were individuals older than 5 years of age with a small number of deceased individuals, an adjustment by single year of age was judged appropriate. The weights of these children were set to zero. In order to compensate for these losses, a simple ratio adjustment by Census single year of age (0, 1, 2, 3, 4 and 5) was done to preserve the total sum of weights in each age group. That is, the weights of the in scope units were inflated such that the sums of the new weights by single year of age were preserved.
7.4.5 Adjustment for non-response
Two adjustments were made for two different types of non-response: the children for whom no contact was made with the parent or guardian (1,700 children) and the children for whom the contacted parent or guardian did not (or could not) provide the information for the child (mainly refusals, about 1,570 children). The weights were first adjusted for non-contact and then for other forms of non-response. In what follows, the term "'non-response" will be used for both types of non-response. The term "respondent" refers to the parent or guardian who is contacted or who provides the information for the selected child once contacted. The term "responding unit" refers to the child for whom a response is obtained from the parent or guardian.
Each non-response adjustment was done in three steps. First, a logistic regression model was used to predict the response probability (probability of obtaining a response) for each child (for both responding and non-responding units) from a series of explanatory variables. These variables, including parent or guardian characteristics, household characteristics and child characteristics, were either Census characteristics, (for example, family structure, Census Aboriginal group of the child) or collection variables (for example, number of attempts to contact a subject, whether field follow-up was required, etc.). As parent or guardian characteristics were required for both responding and non-responding units, the Census family structure was used to determine who would be the "most likely" parent or guardian of the child. Collection variables were found to be particularly good predictors of the response or non-response as many of these variables measure the effort to contact a person or to obtain a response from a contacted person. For instance, individuals requiring a large number of attempts to be contacted were found to be very similar to individuals for whom no contact was made (all attempts failed). The non-response adjustment was then done by forming non-response adjustment classes in such a way that the children in each class had similar response probabilities. Finally, the inverse of the weighted response rate in a class was used as the weighting adjustment factor for that class and the weights of the responding units within the class were adjusted accordingly.
7.4.6 Post-stratification adjustments
The post-stratification adjustment ensures that the sum of the final weights for the responding units matches the population counts from the Census, according to different groups. For ACS, these groups, called post-strata, were defined from combinations of several variables: the Census Aboriginal group (North American Indian, Métis, Inuit, multiple Aboriginal responses or registered Indian / band member only), the Aboriginal population type (identity or ancestry-only), the geographical domain, the province of residence and the age group (0-1, 2-3, 4-5 years old). The weights were adjusted using the ratio of the Census weighted count to the sample weighted count for each post-stratum. This ensures that the sample did not under- or over-represent certain Census Aboriginal groups, regions or age groups.
Since answers to the screening questions (presented in section 5.1.1) can differ between ACS and the Census, a second post-stratification was carried out in the provinces. This guaranteed that the total Aboriginal population (identity or ancestry), as estimated from the ACS filter questions, matched those from the Census filter questions. This post-stratification was done by geographical domain and by age group, according to the total count of Aboriginal people, and not according to each Aboriginal group, in order not to hide the transitions observed between the Census and the survey that are due to such factors as the proxy effect, the time effect and the survey instrument effect (see section 10.0 The relationship between ACS and the Census).
Note that, when individuals are selected with either Aboriginal identity or ancestry, the only possibility is to lose people from the total Aboriginal population since non-Aboriginal persons who could turn out to be Aboriginal persons in the survey were not sampled (except in the territories). This second post-stratification compensates for that fact. However, non-Aboriginal children were sampled in the territories for the purpose of collecting data for the SNC. In the territories, there were proportionally more non-Aboriginal children in the Census who were reported to be Aboriginal children in the ACS than the reverse (Aboriginal children in the Census reported to be non-Aboriginal children in the ACS). Had this second post-stratification been done in the territories, the effect would have been to reduce the weights of the Aboriginal population while they were being increased in the provinces. This is one of the reasons why the second post-stratification was carried out only in the provinces. The other reason is the fact there was no need to post-stratify in the territories since both Aboriginal and non-Aboriginal children were included in the sample (and there was the possibility for non-Aboriginal children to be reported as Aboriginal children in the ACS).
8.0 Data quality
8.1 Sampling errors
The estimates that can be derived from this survey are based on a sample of individuals. Somewhat different estimates might be obtained if a complete census had been taken using the same questionnaire, interviewers, supervisors, processing methods, etc. as those actually used. The difference between an estimate obtained from the sample and the one resulting from a complete count taken under similar conditions is called the sampling error of the estimate.
In order to provide estimates of sampling error for statistics produced in the ACS, a particular type of "bootstrap" method was developed. Several bootstrap methods exist in the literature but none of them was appropriate for the ACS sample design. The particularities of the design that make the estimation of sampling errors difficult are the following:
- Two-phase sample design in which households are selected in the first phase and children in the second phase
- The sampling fraction of the first phase sample (long-form sample) is non-negligible (about 20% in 2B regions) and the second phase sampling fraction is typically quite high in most strata
- The second-phase strata (combinations of domains of estimation, 2B / 2D regions) are non nested within first-phase strata (collection units)
- The method used has to be flexible enough to produce standard statistics such as proportions, totals, means and ratios but also more sophisticated statistics, including percentiles, logistic regression coefficients, etc.
The method developed is a general bootstrap methodology for two-phase sampling (Langlet, E., Beaumont, J.-F. and Lavallée, P., 2008). Several bootstrap methods exist in the literature for one-phase sampling. The most common one is called the "with-replacement" bootstrap and consists of selecting M with-replacement subsamples from the main sample and producing estimates for each subsample. The bootstrap variance estimate (the variance is a particular measure of sampling error) is then calculated as a function of the squared differences between estimates coming from each of the M bootstrap samples and the estimate coming from the survey sample.
The variance calculation is greatly simplified through the use of bootstrap weights. For each subsample, the initial sampling weights first had to be adjusted for bootstrap subsampling which produces what is called "initial bootstrap weights". Since each bootstrap sample is drawn by selecting the units with replacement, a unit can appear several times in a particular bootstrap sample. It can be shown that the bootstrap weights are a function of the initial weight of the observation multiplied by what is called "the multiplicity" of the unit in the bootstrap sample, which is the number of times the unit is selected in the bootstrap sample. The multiplicity of a unit in the bootstrap sample is a random variable following what is called a multinomial distribution. Hence, the bootstrap weights can be seen as the product of the initial sampling weights of the units by a random adjustment factor (in this case, a function of the multiplicity of the unit). Once initial bootstrap weights have been derived, all weight adjustments applied on the initial sampling weights of the full sample are applied to the initial bootstrap weights to obtain the final bootstrap weights which will capture the variance associated with not only the particular sample design but also the variance associated to all weight adjustments applied to the full sample to derive the final weights.
Any bootstrap method can be used by deriving bootstrap weights and any bootstrap weights can be seen as the product of the initial sampling weights and a random adjustment factor. This is the idea of the general bootstrap methodology for two-phase sampling. In the case of a two-phase sample, the variance can be decomposed into two components, each one associated to a phase of sampling. The method generates a random adjustment factor for each phase of sampling. The initial bootstrap weight of a given unit in a bootstrap sample is the product of its initial sampling weight by the values of the two random adjustment factors for that unit.
There is a major advantage of having two sets of random adjustment factors. The first set of adjustment factors can be used for estimates based on the first phase only, that is, estimates based on the Census long-form sample. These estimates are used when the weights are adjusted to the Census totals in the post-stratification adjustment. This will produce variable Census totals from each bootstrap sample and reflects the fact that the Census totals used are based on a sample and are not known fixed totals.
For the Aboriginal Children's Survey (ACS), 1,000 sets of bootstrap weights were generated using the method described above. The method used is slightly biased in the sense that it slightly overestimates the variance. The amount of overestimation was found to be negligible for the ACS. The method can also lead to negative bootstrap weights. To overcome this problem, a transformation was done on the bootstrap weights which reduced their variability. Therefore, the variance calculated on these transformed bootstrap weights has to be multiplied by a factor which is a function of a certain parameter, called phi . The value of the parameter is selected as the smallest integer that makes all bootstrap weights positive. For the ACS, this factor is 3. The variances calculated from the transformed bootstrap weights have to be multiplied by 32 = 9. Alternatively, the CVs obtained (square root of the variance divided by the estimate itself) have to be multiplied by 3. However, most software producing sampling error estimates from bootstrap weights, have an option to specify this adjustment factor, such that the correct variance estimate is obtained without the need of an extra step to multiply by the constant.
It is of course extremely important to use the appropriate multiplicative factor for any estimate of sampling error such as variance, standard error or CV. Omission of this factor would lead to erroneous results and conclusions. This factor is often specified as the "Fay adjustment factor" in software producing sampling error estimates from bootstrap weights.
The measure of sampling error used for the ACS is the coefficient of variation (CV) of the estimate, which is the standard error of the estimate divided by the estimate itself. For this survey, when the CV of an estimate is equal to or higher than 16.6% but smaller than 33.3%, the estimate will be accompanied by the letter "E" to indicate that the data should be used with caution. When the CV of an estimate is equal to or higher than 33.3%, the cell estimate will be replaced by the letter "F" to indicate that the data is suppressed for reasons of reliability. An "X" is used to indicate that an estimate is suppressed to meet the confidentiality requirements of the Statistics Act.
8.2 Non-sampling errors
Errors which are not related to sampling may occur at almost every phase of a survey. Interviewers may misunderstand instructions, respondents may make errors in answering questions, answers may be incorrectly entered on the questionnaire, errors may be introduced in the processing and tabulation of the data and so on. These are all examples of non-sampling errors. Over a large number of observations, randomly occurring errors will have little effect on estimates. However, errors occurring systematically will contribute to biases in the survey estimates.
A pilot test was conducted from January to February 2006 to evaluate the entire survey process, from the questionnaire content to the data processing. This helped reduce the magnitude of non-sampling error.
Coverage errors occur when there are differences between the target population and the sampled population. Because the ACS sample is selected from those who participated in the Census, individuals who did not participate in the Census could not be sampled for the ACS. If this group of individuals is significantly different from the ones who participated in the Census with respect to the characteristics measured in the ACS, a bias could be introduced. This bias is assumed to be relatively small given the very high coverage rate and response rate obtained in the Census and given the adjustments made on the initial Census sampling weights.
Total non-response can be a major source of non-sampling error in surveys depending on the degree to which responding units and non-responding units differ with respect to characteristics of interest. In the ACS, total non-response occurred if no contact was made with the parent or guardian of the selected child or if the contacted parent or guardian did not provide the information for the child. High response rates are essential for quality data. To reduce the number of non-response cases, the interviewers were all trained by Statistics Canada's staff, provided with detailed interviewer manuals, and were under the direction of interviewer supervisors. Refusals were followed up by senior interviewers to encourage respondents to participate in the survey. In regions covered by telephone interviewing, a field follow-up procedure was put in place to further reduce the level of non-response to a minimum.
Partial non-response occurs if the respondent does not answer a specific question, possibly because he or she does not know the answer or the question is felt to be sensitive. Generally, the extent of partial non-response was small in the ACS. Results from the pilot tests were used to evaluate potential problems and changes to the questionnaires were made. In particular, special measures were put in place to facilitate the collection of data from sensitive questions. Where required, special introductions were included (e.g. question on federal residential school), "refused" categories were added, and so on.
A response error occurs when the respondent misunderstands a question or the interviewer records an incorrect answer. Several procedures were taken to minimize this type of error, including interviewer training and qualitative testing of questions.
Processing errors may occur at various stages including coding, data capture and editing. Quality control procedures were applied to every stage of the data processing to minimize this type of error.
9.1 Analytical files
Accompanying the release of data from the Aboriginal Children's Survey was an analytical article entitled "Aboriginal Children's Survey 2006: Family, Community and Child Care", which provides information on the families, communities, cultural activities and child care arrangements for off-reserve First Nations, Métis and Inuit children, under 6 years of age. The findings are presented separately for each Aboriginal group. Fact sheets, providing highlights from the analytical article, are also available.
9.2 Data products and services
The master data file for the 2006 ACS is available in Statistics Canada's Research Data Centres (RDCs). Accompanying the file, is the record layout, SAS and SPSS syntax to load the file, as well as metadata in the form of a codebook that describes each variable and provides weighted and unweighted frequency counts.
Supporting data tables that provide provincial and territorial estimates, as well estimates for Inuit regions, for key indicators from the released analytical article are available.
Profiles that provide information on a variety of topics covered in the ACS are available on Statistics Canada's website. Information is displayed for different concepts and levels of geography.
A remote data access service will be available in the spring of 2009. This service will allow researchers to write and test programs on a "synthetic file", which has the same layout as the ACS master data file, however it does not contain the real information from respondents. Once programs have been written and tested, researchers can submit their programs to Statistics Canada, who will run the programs using the master data file, vet the results for confidentiality and return the aggregate data tables to the researchers. This service will be offered on a cost-recoverable basis.
Custom tabulations will be produced, upon request, on a cost-recoverable basis.
9.3 Survey documentation
Information about the Aboriginal Children's Survey is available on Statistics Canada's website. This information includes:
- Concepts and Methods Guide
- User's Guide
- Integrated Metadatabase (IMDB)
10.0 The Relationship between the Aboriginal Children's Survey and the Census
As discussed above, the ACS is a post-censal survey, which means that Census information was used to determine who would be included in the ACS sample. More detailed information about how Census responses were used to determine the population of interest for the ACS is provided in Section 5.0 (Survey Design).
The Census and the ACS are both rich sources of information on Aboriginal peoples that complement each other. The ACS takes concepts that are touched on in the Census and asks questions that dig deeper in order to provide more detailed information. For example, the Census provides some information on the language spoken most often at home, mother tongue and which languages the child can speak. Adding information from the ACS provides the opportunity to learn how often the child is exposed to an Aboriginal language, in which languages the child can express his or her needs, etc.
The ACS also covers entire topics or themes that are not included in the Census. For example, the ACS can provide extensive information on child health, nutrition, sleep and early childhood development among others.
Both the Census and the ACS conceptually cover the two types of Aboriginal populations, that is, the "identity population" and the "ancestry population" as described in section 5.
10.1 Differences in counts
While the post-stratification (see Section 7.4.6) ensured that the total number of people with Aboriginal ancestry or identity is the same for the Census and the ACS (except in the territories due to the sampling of the non-Aboriginal population), it did not ensure that the counts for the Aboriginal groups would match. Indeed, the Census and the ACS produce different counts at the Aboriginal group level. This is due to changes in the way respondents answered questions about their Aboriginal ancestry and Aboriginal identity from the time of the Census to the time of the ACS. Respondents may have changed their responses for a number of reasons, including differences in how the information was collected.
10.1.1 Different modes of interview
Most of the 2006 Census data were collected through self-enumeration using a mail-out mail-back methodology (except for Indian reserves and remote areas, including all Inuit communities, where a canvasser methodology was used). In general, one household member completed the Census form for all household members. This is called proxy reporting, meaning someone other than the person for whom the information is reported answers the questions.
In all Inuit communities, all of the Northwest Territories (except Yellowknife) and Labrador, ACS data were collected through personal interviews. Everywhere else in Canada, data were collected mostly though telephone interviews (some places had field follow-up done at the end of data collection to reduce non-response). In the ACS, the interview is completed by one of the child's parents or guardians. Because the person contacted for the ACS may not be the same person who filled in the Census questionnaire, there may be some differences in responses.
10.1.2 Different questionnaires
Another source of discrepancy between the Census and the ACS is the "ethnic origin" or "ancestry" question. The Census uses an open-ended ethnic origin question (To which ethnic or cultural group(s) did this person's ancestors belong?) Answers to this write-in question are coded to determine whether the person has Aboriginal ancestry, and, if they do, which Aboriginal ancestry group(s) they fall into (North American Indian, Métis and / or Inuit). In the Aboriginal Children's Survey (ACS), three Aboriginal group-specific questions are asked regarding North American Indian, Métis and Inuit ancestries.
As a result, more of the respondents to the ACS reported Aboriginal ancestry in the ACS than in the Census, with many more multiple combinations. For example, one may have written in "Métis" on the Census ancestry question, and then reported having both North American Indian and Métis ancestries when asked about each group in the ACS.
The Aboriginal self-reporting question is essentially the same on both the Census and the ACS forms (Is ____ an Aboriginal person, that is, North American Indian, Métis or Inuit?). However, on the Census form, there is an instruction saying "If "yes", mark "x" the circle(s) that best describe(s) this person now". This may influence the respondent to choose the category that best describes the person concerned, and therefore mark only one category as opposed to many. During the ACS training, interviewers were asked to pay attention to the possibility of having multiple Aboriginal self-reporting and to read the question completely, including the three Aboriginal groups. This may have led to the reporting of more Aboriginal groups in the ACS than in the Census. Also, because of the fact that in the ACS, the Aboriginal self-reporting question is preceded by three specific questions on Aboriginal ancestries (three questions in one) and not by a general open-ended ethnic origin question as in the Census, respondents may be more likely to report themselves as an Aboriginal person with the ACS question.
10.1.3 Different context
The Census form is very general in terms of content whereas the ACS is a survey specifically designed for Aboriginal people. As a result, individuals may have given more detailed information about their Aboriginal ancestry and Aboriginal identity in the ACS.
10.1.4 Coverage and sampling methodology
Except in the territories (where the non-Aboriginal population was also sampled), the ACS sample was selected among those who reported Aboriginal ancestry and / or Aboriginal identity in the Census. However, when contacted for the ACS, some individuals no longer reported their child as having Aboriginal ancestry or Aboriginal identity. This may have been due to several factors. In particular, in the case of a father with Aboriginal identity and a non-Aboriginal mother, the father may very well report an Aboriginal identity for his child on the Census form whereas the mother may not have reported it in the ACS when contacted. As mentioned above, in order to compensate for any such loss in the overall Aboriginal population, a post-stratification adjustment was carried out as part of the weighting process.
It is important to note that there were substantial transitions between the Aboriginal ancestry population and the Aboriginal identity population from the time of the Census to the ACS. Some individuals who reported Aboriginal identity for their child in the Census reported only Aboriginal ancestry for their child (with no Aboriginal identity) in the ACS. Conversely, some individuals who reported only Aboriginal ancestry for their child (with no Aboriginal identity) in the Census reported an Aboriginal identity for their child in the ACS. For the reasons described above, a larger group of individuals fell into the second category – in other words many children "gained" Aboriginal identity in the ACS. As a result of this effect, the count of the total child Aboriginal identity population will be larger from the ACS than from the Census. The count of children with only Aboriginal ancestry (with no Aboriginal identity) will be smaller from the ACS than from the Census.
An example to illustrate how one may move from having only an Aboriginal ancestry (with no Aboriginal identity) in the Census to the Aboriginal identity population in the ACS may help clarify this. On the Census, a person reports his or her child having a North American Indian ancestry (along with non-Aboriginal ancestries such as Irish and Scottish), but does not report the child as having Aboriginal identity. When contacted for the ACS, the same person or his or her spouse reports the child as having North American Indian ancestry and North American Indian identity. This means that the child has moved from the Aboriginal ancestry-only population in the Census (and therefore not being counted in the identity population) to the Aboriginal identity population for the ACS.
On the other hand, because of the Aboriginal group-specific nature of the ancestry question on the ACS, the number of children being reported as having Aboriginal identity only (with no Aboriginal ancestry) is substantially smaller for the ACS than the Census. For example, on the Census, some individuals may report their children as having just a French ancestry with a Métis identity. When contacted for the ACS, they may have been more specific about the ancestry of their children. They may have reported them as having both North American Indian and French ancestries in addition to reporting them as having a Métis identity (It is common for Métis people to have North American Indian and French ancestries). They have then moved from having only Aboriginal identity on the Census, to having both Aboriginal identity and Aboriginal ancestry on the ACS.
Transitions between the different Aboriginal groups (North American Indian, Métis and Inuit) also occurred. For example, one may have reported his or her child as having a North American Indian identity on the Census, but both a North American Indian and a Métis identity in the ACS.
The following tables compare the Census counts to the ACS counts for different geographical regions and Aboriginal groups. For these tables, the non-Aboriginal population covered in the ACS for the territories is excluded. The four Inuit regions are separated from the rest of Canada. Hence, the region Atlantic Other excludes Nunatsiavut, Quebec Other excludes Nunavik and Northwest Territories Other excludes Inuvialuit.
Tables 3 and Table 4 compare respectively the non-reserve (except in the territories where First Nations communities are included) Census and the ACS counts for the identity population and ancestry-only population without double counting.
Tables 5 and 6 are similar to Tables 3 and 4, but include double counting. This means that someone with a multiple identity of NAI and Métis counts in both the NAI and Métis categories.
Identity counts for the Census and the Aboriginal Children's Survey without double counting, 2006
Ancestry-only counts for the Census and the Aboriginal Children's Survey without double counting, 2006
Identity counts for the Census and the Aboriginal Children's Survey with double counting, 2006
Ancestry-only counts for the Census and the Aboriginal Children's Survey with double counting, 2006
- The terms First Nations and North American Indian are used interchangeably throughout this document
- The First Nations communities covered in Quebec have been removed from the analytical file to facilitate representative analysis of the Aboriginal Children's Survey (ACS) The remainder of this document highlights the concepts and methods used in producing the data that appears on the analytical file.
- Date modified: