Canadian Housing Survey

Detailed information for 2021

This survey collects information about housing needs and experiences from a sample of Canadian households. Information is collected on housing conditions and costs, dwelling and neighbourhood satisfaction, housing moves, and other aspects of well-being related to housing. Collected information allows for the derivation of key housing indicators such as core housing need.

Data release - July 21, 2022

• Questionnaire(s) and reporting guide(s)
• Description
• Data sources and methodology
• Data accuracy

Description

The Canadian Housing Survey (CHS) provides information on how Canadians feel about their housing and how housing affects them. Information is collected on housing conditions and costs; dwelling characteristics and housing tenure; perceptions of economic hardship from housing costs; dwelling and neighbourhood satisfaction; perceptions of neighbourhood issues and safety; housing moves including forced moves; civic engagement; life and community satisfaction; self-assessed health; various dimensions of physical and mental well-being; experience with homelessness; socio-demographic characteristics; and the impacts of COVID-19 on some aspects of housing.

Social and Affordable Housing (SAH) is an important part of the housing stock that provides housing to people in need. The CHS aims to provide detailed and precise statistics on households in SAH by oversampling this subpopulation.

The 2021 CHS was carried out in all ten provinces and the three territorial capitals.

Subjects

• Families, households and housing
• Household characteristics

Data sources and methodology

Target population

The target population is the population of Canada's 10 provinces excluding residents of institutions, members of the Canadian Forces living in military camps and people living on Indian reserves. In all, these exclusions make up about 2% of the population.

People living in other types of collective dwellings are also excluded from the survey:
- people living in residences for dependent seniors; and
- people living permanently in school residences, work camps, etc.; and
- members of religious and other communal colonies.

Collection exclusions make up less than 0.5% of the target population. However, these people are included in the population estimates to which the CHS estimates are adjusted.

Instrument design

The questionnaire content was determined in consultation with the survey sponsor (Canada Mortgage and Housing Corporation). Standardized content was used when available. Harmonized content was also re-used from past or existing surveys. Questionnaire design specialists were consulted during the design phase. The survey also underwent qualitative testing prior to collection, and recommendations were incorporated.

The questionnaire follows standard practices and wording used in electronic questionnaires, such as the automatic control of flows that depend on answers to earlier questions and the use of edits to check for logical inconsistencies and input errors. The computer application for data collection was tested extensively.

Sampling

Sampling unit:
The sampling unit for the CHS is the private dwelling.

Stratification method:
The 2021 CHS uses the Dwelling Universe File (DUF) as the survey frame. Administrative data is used to classify SAH dwellings into separate strata on the frame. The frame is also stratified into geographic areas of interest based on census subdivision (CSD) boundaries.

There are 43 different geographic strata including the largest census metropolitan areas (CMAs) in each province; CMAs with a population of more than 500,000 according to the 2016 Census; the census agglomerations (CAs) of Charlottetown, Yellowknife and Whitehorse, and the census subdivision (CSD) of Iqaluit; combined CMAs and combined CAs in each province; as well as the regions outside CMAs and CAs in each province.

Each geographic stratum is divided into two groups: SAH dwellings and all other dwellings. Sub-strata are used to more efficiently sample rented and owned dwellings within the non-SAH strata.

Sampling and sub-sampling:
The CHS is cross-sectional with dwellings selected at random for each new survey cycle.

The CHS oversamples SAH dwellings. As a result, the sample includes a higher proportion of SAH dwellings than what is present in the housing stock (about 25% in the sample versus around 5% in the housing stock).

SAH dwellings are an important part of the housing stock that provides housing at below-market rents to households in need. The oversampling of SAH dwelling helps to ensure there will be enough survey responses to produce accurate statistics for key housing variables in most CHS strata for this sub-population.

The sample size for the CHS is determined by calculating the number of respondents necessary to create estimates with the desired level of precision in each stratum. Precision is measured by the coefficient of variation (CV). The target CV is 10% in CMAs and CAs; 15% outside CMAs and CAs and in the territories; and 15% in the SAH strata. The sample size for the 2021 CHS is about 95,800 dwellings, of which about 25,100 are SAH.

Within each stratum (geography by SAH status and tenure), a systematic random sample is independently selected. Records are sorted by predicted household income, obtained from administrative records. This ensures the sample includes households with low, medium and high income, making the sample more representative.

Data sources

Data collection for this reference period: 2021-01-04 to 2021-06-02

Responding to this survey is voluntary.

Data are collected directly from survey respondents and extracted from administrative files.

The CHS is a biennial survey. Results from its first cycle were released in 2018. Due to the global pandemic, the launch of collection for cycle two was delayed from November 2020 to January 2021. All data collected for cycle two from respondents were reported for reference months in 2021. As a result, the cycle 2 reference period is 2021 instead of 2020 as originally planned. However, the income data from administrative tax files linked to the 2021 CHS are for reference year 2020 since these data are only available with a one-year lag. This means total before-tax household income and income thresholds for 2020 were used to calculate the CHS 2021 measures of core housing need.

For the 2021 CHS, data were collected using two collection modes: self-response Electronic Questionnaire (rEQ) and Computer Assisted Telephone Interviewing (CATI). Invitation letters to participate in the 2021 CHS were mailed to sampled households. Letters provided a link to the survey and a unique secure access code. Households that did not participate via rEQ were contacted by CATI interviewers for follow-up.

The CHS asked that the survey be completed by the household member with the most knowledge of the household's housing situation. In all cases, this person was aged 15 years or older. Proxy response was accepted for questions about other household members. This allowed one household member to answer questions on behalf of any or all other household members.

View the Questionnaire(s) and reporting guide(s) .

Error detection

The computerized questionnaire contains many features designed to maximize the quality of the collected data. Edits are built into the questionnaire to compare the reported data with unusual values and to detect logical inconsistencies. When an edit fails, the interviewer is prompted to correct the information (with the respondent's help, if necessary). Once the data are transmitted to Head Office, a comprehensive series of processing steps is undertaken for the detailed verification of each completed questionnaire. Invalid responses are corrected or flagged for imputation.

Imputation

Individuals from a household were considered respondents if they met certain criteria after providing sufficient demographic information about themselves. A household was considered a respondent if at least one person in the household met the criteria. Any missing key data—either person-level data for individuals within the responding household, or data for household-level variables—were imputed.

Missing responses were imputed for key questionnaire fields in two stages. First, imputation was completed for demographic variables at the person level (e.g., age, sex and relationships). Second, household-level variables were imputed (e.g., ownership, need of repairs and rent subsidy).

The CHS uses the nearest-neighbour imputation method, which involves selecting a donor record based on matching variables. The imputation is performed using Statistics Canada's Census Edit and Imputation System (CANCEIS). First, a set of matching variables is defined, where each variable is correlated with the variables to be imputed. Second, through the combined use of a score function (for categorical matching variables) and a distance function (for numeric matching variables), the most similar consistent donor record is identified and used to impute data for the record with missing fields. For the mortgage questions, some deterministic imputation was done based on respondent comments provided in the questionnaire.

Not all variables in the CHS database were imputed for non-response.

Estimation

An integrated weight was produced for the CHS. This means that all household members were given the same weight. The weighting process begins by calculating design weights, which are the inverse of the probability of selection. The design weights are the same for all units within the same geographic and tenure strata. The design weights were then adjusted to compensate for non-response. Weights were adjusted separately within groups with similar response propensity so that the respondents within the group also represented the non-respondents. The adjustment groups were formed using auxiliary variables, which were available for all sampled units, were associated with non-response and were related to the survey's key variables of interest. After the non-response adjustment, extreme weights, generally caused by misclassification of the SAH and tenure strata, were identified and reduced for a small number of observations. The weights were then post-stratified to ensure they matched the preliminary 2021 Census estimates for tenure and subsidized dwellings.

Next, the weights underwent an initial calibration to ensure that certain weighted estimates respected relevant population totals from reliable sources. For each province, population counts by age and sex groups and household size were used. The CHS also used population counts for the largest CMAs targeted by the survey. The second set of totals was the number of wage and salary earners by province. The population totals used for the CHS included demographic projections produced by Statistics Canada's Demography Division (based on the 2016 Census) and the number of wage and salary earners by province (based on the Canada Revenue Agency's T4 file of 2020). The CHS was also calibrated to three dwelling types: single detached dwellings, apartment buildings of fewer than five storeys and apartment buildings of five or more storeys.

To complete the weighting process, outlier weight adjustment and calibration steps were conducted a second time. The resulting weighted units represented the entire 2021 CHS target population.

To estimate sampling variance, the Rao-Wu-Yue bootstrap approach was used. To determine the bootstrap weights, 1,000 initial replicates were created, and then each replicate underwent the same adjustment process as the survey weights.

Quality evaluation

When all processing, weighting and estimation steps were complete, CHS estimates for key housing and related variables were compared with other data sources such as the Census, administrative sources and other Statistics Canada surveys.

Disclosure control

Statistics Canada is prohibited by law from releasing any information it collects which 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.

Revisions and seasonal adjustment

Revisions are made to the CHS when new population estimates become available following the most recent census. At that time, CHS data based on projections from the previous census are re-weighted using the new population estimates, and the corresponding historical CHS estimates are revised. With the release of the 2021 CHS, estimates from the 2018 CHS have been revised to reflect population projections based on the 2016 Census.

Data accuracy

Response rates
The overall response rate for the 2021 CHS was 47%.

All surveys are subject to sampling and non-sampling errors. Sampling errors occur because inferences about the entire population are based on information obtained from only a sample of the population. The sample design, estimation method, sample size and data variability determine the size of the sampling error. A common measure of sampling errors is a Confidence Interval (CI), where wider CIs indicate lower precision of survey estimates.

Non-sampling errors occur because certain factors make it difficult to obtain accurate responses or to ensure that responses retain their accuracy throughout processing. Unlike sampling errors, non-sampling errors are not easily quantified. Four types of non-sampling errors can be identified: coverage errors, response errors, processing errors and non-response errors. Coverage errors arise when sampling frame units do not adequately represent the target population.

Response errors occur when respondents provide inaccurate information. Processing errors may occur in any of the data processing stages, including data entry, coding, editing, imputation of partial non-response, weighting and tabulation.

Non-response errors occur when potential respondents do not provide the required information, or when the information they provide is unusable. The main impact of non-response on data quality is that it can cause bias in the estimates if the characteristics of non-respondents differ from those of respondents in a way that affects the information collected by the survey. Although non-response rates can be calculated, they only provide an indication of data quality since they do not measure the degree of bias present in the estimates. The magnitude of the non-response rate can be considered a simple indicator of the risk of bias in the estimates.

While the weights of respondent households are adjusted to compensate for non-respondent households, partial non-response (failure to answer some questions) is handled through imputation. Non-response weight adjustment and imputation help mitigate the impact of non-response on survey estimates.