General Social Survey - Family (GSS)
Detailed information for 2017 (Cycle 31)
Status:
Active
Frequency:
Every 5 years
Record number:
4501
The two primary objectives of the General Social Survey are to gather data on social trends in order to monitor changes in the living conditions and well-being of Canadians over time, and to provide information on specific social policy issues of current or emerging interest.
Data release - February 7, 2019
- Questionnaire(s) and reporting guide(s)
- Description
- Data sources and methodology
- Data accuracy
- Documentation
Description
The General Social Survey - Family monitors changes in Canadian families. It collects information on conjugal and parental history (chronology of marriages, common-law unions and children), family origins, children's home leaving, fertility intentions, and other socioeconomic characteristics.
The information collected will impact program and policy areas such as parental benefits, child care strategies, child custody and spousal support programs.
Statistical activity
This record is part of the General Social Survey (GSS) program. The GSS originated in 1985. Each survey cycle contains a core topic, focus or exploratory questions and a standard set of socio-demographic questions used for classification. More recent cycles have also included some qualitative questions, which explore intentions and perceptions.
Subjects
- Education, training and learning
- Families, households and housing
- Health
- Immigration and ethnocultural diversity
- Indigenous peoples
Data sources and methodology
Target population
The target population for the 2017 General Social Survey is all non-institutionalized persons 15 years of age or older, living in the 10 provinces of Canada.
Instrument design
The questionnaire was designed based on research and extensive consultations with key partners and data users. Qualitative testing, conducted by Statistics Canada's Questionnaire Design Resource Center (QDRC), was carried out, with respondents in two cities, who were screened in based on representative criteria. Questions which worked well and others that needed clarification or redesign were highlighted. QDRC staff compiled a detailed report of the results along with their recommendations. All comments and feedback from qualitative testing were carefully considered and incorporated into the survey whenever possible.
Sampling
This is a sample survey with a cross-sectional design.
This survey uses a frame that combines landline and cellular telephone numbers from the Census and various administrative sources with Statistics Canada's dwelling frame. Records on the frame are groups of one or several telephone numbers associated with the same address (or single telephone number in the case a link between a telephone number and an address could not be established). This sampling frame is used to obtain a better coverage of households with a telephone number.
The sample is based on a stratified design employing probability sampling. The stratification is done at the province/census metropolitan area (CMA) level. Information is collected from one randomly selected household member aged 15 or older, and proxy responses are not permitted.
Sampling Unit
The General Social Survey (GSS) uses a two-stage sampling design. The sampling units are the groups of telephone numbers. The final stage units are individuals within the identified households. Note that GSS only selects one eligible person per household to be interviewed.
Stratification method
In order to carry out sampling, the 10 provinces of the target population are divided into strata (i.e. geographic areas). Many of the CMAs are each considered separate strata. This was the case for St. John's, Halifax, Saint John, Montreal, Quebec City, Toronto, Ottawa (Ontario part of Ottawa-Gatineau CMA), Hamilton, Winnipeg, Regina, Saskatoon, Calgary, Edmonton, Vancouver and Victoria. CMAs not on this list are located in New Brunswick, Quebec, Ontario and British Columbia. For Quebec, Ontario and British Columbia, three more strata were formed by grouping the remaining CMAs in each of these three provinces (Québec part of Ottawa-Gatineau CMA of is in Quebec-Other-CMAs). Next, the non-CMA areas of each of the 10 provinces were grouped to form 10 more strata. Moncton is included with the non-CMA group for New Brunswick. This resulted in 27 strata in all.
Sampling and sub-sampling
For each province, minimum sample sizes were determined that would ensure certain estimates would have acceptable sampling variability at the stratum level. Once these stratum sample size targets had been met, the remaining sample was allocated to the strata in a way that balanced the need for precision of both national-level and stratum-level estimates. This sample was representative of all households in the 10 provinces.
For the survey, a single eligible member of each sampled household is randomly selected by the application to complete the questionnaire, after the completion of the roster.
A field sample of approximatively 43,000 units was used. Among them, about 34,000 invitation letters were sent to selected households across Canada. A completion of 20,000 questionnaires was expected.
Data sources
Data collection for this reference period: 2017-02-01 to 2017-11-30
Responding to this survey is voluntary.
Data were collected directly from survey respondents using the computer-assisted telephone interviewing method. No proxy reporting was allowed. Respondents could be interviewed in French or English.
Tax derived files (Canadian Statistical Demographic Database environment).
Questions relating to income show rather high non-response rates, the incomes reported by respondents are usually rough estimates. Linking allows getting such information without having to ask questions.
The information collected during the 2017 General Social Survey (Cycle 31) has been linked to the personal tax records (T1 Family File or T4 Information Slip) of respondents. Household information (address, postal code, and telephone number) and respondent's information (social insurance number, surname, name, date of birth/age and sex) are key variables for the linkage.
Respondents were notified of the planned linkage before and during the survey. Any respondents who objected to the linkage of their data had their objections recorded, and no linkage to their tax data took place.
View the Questionnaire(s) and reporting guide(s) .
Error detection
Processing used the Social Survey Processing Environment set of generalized processes to allow subject matter and survey support staff to specify and run the processing of the survey in a timely fashion with high quality outputs.
It used a structured environment to monitor the processing of data ensuring best practices and harmonized business processes were followed.
Edits were performed automatically and manually at various stages of processing at macro and micro levels. They included family, consistency and flow edits. Family relationships were checked to ensure the integrity of matrix data. A series of checks were done to ensure the consistency of survey data. An example was to check the respondent age against the respondent birth date. Flow edits were used to ensure respondents followed the correct path and fix off-path situations. Error detection was done through edits programmed into the computer-assisted telephone interviewing (CATI) system.
The CATI data capture program allowed a valid range of codes for each question and built-in edits, and automatically follows the flow of the questionnaire.
All survey records were subjected to computer edits throughout the course of the interview. The CATI system principally edited the flow of the questionnaire and identified out of range values. As a result, such problems were immediately resolved with the respondent. If the interviewer was unable to correctly resolve the detected errors, the interviewer bypassed the edit and forwarded the data to head office for resolution. All interviewer comments were reviewed and taken into account by head office editing.
Head office performed the same checks as the CATI system as well as the more detailed edits discussed previously.
Imputation
In 2017, personal income questions were not asked as part of the survey. Income information was obtained instead through a linkage to tax data for respondents who did not object to this linkage. Income information was obtained from the 2016 T1 Family File for 83.1% of the respondents. Missing information for all other respondents was imputed. Since the 2016 General Social Survey (GSS), the family income (i.e., linking directly to a variable on the T1 Family File that corresponds to the census family income) is used instead of the household income. In total, a family income value was obtained for 82.6% of households for GSS 2017.
Estimation
When a probability sample is used, as was the case for this survey, the principle behind estimation is that each person selected in the sample represents (in addition to himself/herself) several other persons not in the sample. For example, in a simple random sample of 2% of the population, each person in the sample represents 50 persons in the population (himself/herself and 49 others). The number of persons represented by a given respondent is usually known as the weight or weighting factor.
The 2017 General Social Survey is a survey of individuals and the analytic files contain questionnaire responses and associated information from the respondents.
A weighting factor is available on the microdata file:
WGHT_PER: This is the basic weighting factor for analysis at the person level, i.e. to calculate estimates of the number of persons (non-institutionalized and aged 15 or over) having one or several given characteristics.
In addition to the estimation weights, bootstrap weights have been created for the purpose of design-based variance estimation.
Estimates based on the survey data are also adjusted (by weighting) so that they are representative of the target population with regard to certain characteristics (each month we have independent estimates for various age-sex groups by province). To the extent that the characteristics are correlated with those independent estimates, this adjustment can improve the precision of estimates.
Quality evaluation
While rigorous quality assurance mechanisms are applied across all steps of the statistical process, validation and scrutiny of the data by statisticians are the ultimate quality checks prior to dissemination. Many validation measures were implemented. They include:
a) analysis of changes over time
b) verification of estimates through cross-tabulations
c) confrontation with other similar sources of data.
Disclosure control
Statistics Canada is prohibited by law from releasing any information it collects that could identify any person, business, or organization, unless consent has been given by the respondent or as permitted by the Statistics Act. Various confidentiality rules are applied to all data that are released or published to prevent the publication or disclosure of any information deemed confidential. If necessary, data are suppressed to prevent direct or residual disclosure of identifiable data.
In order to prevent any data disclosure, confidentiality analysis is done using the Statistics Canada Generalized Disclosure Control System (G-Confid). G-Confid is used for primary suppression (direct disclosure) as well as for secondary suppression (residual disclosure). Direct disclosure occurs when the value in a tabulation cell is composed of or dominated by few enterprises while residual disclosure occurs when confidential information can be derived indirectly by piecing together information from different sources or data series.
Revisions and seasonal adjustment
This methodology does not apply to this survey program.
Data accuracy
As the data are based on a sample of persons, they are subject to sampling error. That is, estimates based on a sample will vary from sample to sample, and typically they will be different from the results that would have been obtained from a complete census. More precise estimates of the sampling variability of estimates can be produced with the bootstrap method using bootstrap weights that have been created for this survey. The bootstrap method was used to estimate the sampling variability for all of the estimates produced based on the data from the 2017 General Social Survey (GSS). Estimates with high sampling variability are indicated in this publication and all of the highlighted differences between subgroups of the population are significant at the 95% level.
Response rate:
The overall response rate is 52.4%.
Non-sampling error:
Common sources of these errors are imperfect coverage and non-response. Coverage errors (or imperfect coverage) arise when there are differences between the target population and the surveyed population. Households without telephones, as well as households with telephone services not covered by the current frame, represent a part of the target population that was excluded from the surveyed population. To the extent that the excluded population differs from the rest of the target population, the results may be biased. In general, since these exclusions are small, one would expect the biases introduced to be small. Non-response could occur at several stages in this survey. There were two stages of information collection: at the household level and at the individual level. Some non-response occurred at the household level, and some at the individual level. Survey estimates will be adjusted (i.e. weighted) to account for non-response cases. Other types of non-sampling errors can include response errors and processing errors.
Non-response bias:
The main method used to reduce nonresponse bias involved a series of adjustments to the survey weights to account for nonresponse as much as possible. For the 2017 GSS, an additional adjustment was added where basic characteristics of non-responding households, such as income and household composition, were extracted from the 2016 Census data and then used to model and adjust nonresponse.
Coverage error:
The frame for GSS was created using several linked sources, such as the Census, administrative data and billing files. Coverage was improved (overcoverage and undercoverage may still exist) if we compare it to the random digit dialling strategies used in the past. All respondents in the 10 provinces were interviewed by telephone. Households without telephones were therefore excluded from the survey population. Survey estimates were adjusted (weighted) to represent all persons in the target population, including those not covered by the survey frame.
Other non-sampling errors:
For the 2017 GSS, significant effort was made to minimize bias by using a well-tested questionnaire, a proven methodology, specialized interviewers and strict quality control.
Documentation
- The General Social Survey: An Overview
Last review : January 7, 2021
- Date modified: