Canadian Survey of Economic Well-being (CSEW)
Detailed information for 2013
The purpose of the survey is to provide information to the Government of Canada on the economic well-being of Canadian households. The data will be used to assess the needs of households in Canada.
Data release - December 12, 2013
The objective of the survey is to produce estimates of the incidence of goods or activities that are lacking at the national and provincial geographies, at the household level to examine the basic economic needs and financial circumstances of households in Canada. The information will be used by Employment and Social Development Canada, Statistics Canada and other research organizations interested in improving well-being.
- Household, family and personal income
- Income, pensions, spending and wealth
- Low income and inequality
Data sources and methodology
The target population for this survey consists of all individuals in Canada, excluding residents of the Yukon, the Northwest Territories and Nunavut and residents of institutions and persons living on Indian reserves. Overall, these groups together represent an exclusion of less than 3% of the population.
Data for this survey were collected using a questionnaire of approximately 40 questions. Experts from Employment and Social Development Canada and from Statistics Canada worked closely together to develop this collection instrument. The questionnaire underwent an in-depth review by a multidisciplinary team from Statistics Canada and was then tested thoroughly, in collaboration with the agency's Questionnaire Design Resource Centre, using cognitive interviews. The questionnaire was finalized in June 2013.
This is a sample survey with a cross-sectional design.
The questionnaire is administered to a sub-sample of the individuals already selected for the Labour Force Survey (LFS), record number 3701. The LFS sample is drawn from an area frame and is based on a stratified, multi-stage design that uses probability sampling. The total LFS sample is composed of six independent samples, called rotation groups, because each month one sixth of the sample (or one rotation group) is replaced.
CSEW uses four rotation groups from the LFS (sample size is about 8,000 per rotation group).
Data collection for this reference period: 2013-08-18 to 2013-10-07
Responding to this survey is voluntary.
Data are collected directly from survey respondents.
The LFS interview is completed over the phone by a responsible member of the household who generally provides the LFS responses for all members. Following the LFS interview, and subject to operational constraints, the interviewer then requests that this same member answers the CSEW questionnaire for all members of the household.
Interviews are conducted from Statistics Canada's regional offices using a Computer Assisted Telephone Interviewing (CATI) application.
View the Questionnaire(s) and reporting guide(s) .
Editing for this survey was done directly at the time of the interview. Where the information entered is out of range (too large or small) of expected values, or inconsistent with the previous entries, the interviewer was prompted, through message screens on the computer, to modify the information.
Donor imputation was used to fill in missing data for non-response to the questions related to income. Some respondents did not want or were not able to report salary or personal income other than by selecting a range. In those cases too, donor imputation was used to assign a corresponding dollar value to the unit. Donor imputation involved donors that were selected using a score function. For each item non-response or partial non-response records (also called recipient records), certain characteristics were compared to characteristics from all the donors. When the characteristics were the same between a donor and the recipient, a value was added to the score of that donor. The donor with the highest score was deemed the "closest" donor and was chosen to fill in missing pieces of information of the non-respondents. If there was more than one donor with the highest score, a random selection occurred. The pool of donors was made up in such a way that the imputed value assigned to the recipient, in conjunction with other non-imputed items from the recipient would still pass the edit.
There were a few high-income households who were excluded from the donor pool. Their reported income examined in the context of demographic and labour force characteristics was unique, so they did not meet the criteria of a good donor.
To reduce response burden, personal income was collected for a maximum of four households members aged 15 and over. There were around 500 individuals that belonged to households with five adults or more. These individuals were processed as other non-respondents and had their income value imputed.
The CSEW sample is a sub-sample selected from the Labour Force Survey (LFS) sample. LFS uses a complex random sampling plan to select the households. Each household in the sample represents a number of other households in the population. Estimates for a given characteristic are obtained by multiplying the household weight by the corresponding value of the characteristic for the household. The key step in the point estimation process is therefore the derivation of the weight which is explained next.
We start from the LFS weight. Then, we make an adjustment to that weight to account for the fact that the CSEW is a sub-sample of the LFS sample.
Then, an adjustment to account for non-response is made (non-response to the CSEW households that responded to the LFS). This weight adjustment is made within what is known as response homogeneity groups defined as combination of key demographic variables where households within groups are deemed homogenous to each other with respect to the characteristics observed (income and well-being variables). Next, we use a statistical "weight share" technique to derive person level weights while maintaining a household weight. This gives the flexibility to data users to conduct micro analyses at the person and/or at the household level.
Independent estimates are also available monthly for various age and sex groups by province. Those are population projections based on the most recent census data, records of births and deaths, and estimates of migration. Using a regression model, we derive a final set of weights using the projections as benchmarks. The set up is such that the final weights produced will sum exactly to the census projections, improving thus the reliability of the estimates produced by the CSEW.
To provide an estimate of the accuracy of the data, the coefficient of variation is usually the measure retained. This implies that an estimate of standard deviation must be obtained. Given the complexity of the sampling plan, direct formulas to calculate the standard deviation cannot be derived. A replication technique, namely the bootstrap method, is therefore used instead.
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.
Revisions and seasonal adjustment
This methodology does not apply to this survey.
Accuracy of the data is affected by all the sources of survey errors. They can be classified into two categories: non-sampling errors and sampling error.
Non-sampling errors are errors arising during the course of virtually all survey activities, apart from sampling. They arise primarily from the following sources: non-response, coverage, measurement and processing.
Non-response errors result from a failure to collect complete information on all units in the sample. Non-response produces errors in the survey estimates in two ways. First, is that non-respondents often have different characteristics from respondents; secondly, it reduces the effective size of the sample, since fewer units than expected answered the survey. As a result, the sampling variance increases and the precision of the estimates decrease. While the LFS response rate is around 92% at Canada level and provincially, the CSEW response rate varies provincially from 63% to 72% and is 67% for the whole country.
Coverage errors consist of omissions, erroneous inclusions, duplications and misclassifications of units in the survey frame. Since they affect every estimate produced by the survey, they are one of the most important types of error. Coverage errors may cause a bias in the estimates and the effect can vary for different sub-groups of the population. The LFS frame excludes less than 3% of all households in the 10 provinces of Canada. Therefore, the CSEW frame also excludes a similar proportion of households. It is unlikely that this exclusion introduces any significant bias.
Measurement errors occur when the response provided differs from the real value; such errors may be attributable to the respondent, the interviewer, the questionnaire, the collection method or the respondent's record-keeping system. Such errors may be random or systematic. Several measures are taken to reduce the level of response error: questionnaire review and testing, the use of highly skilled interviewers, extensive training of interviewers, observation and monitoring of interviewers to detect problems of questionnaire design or misunderstanding of instructions.
Processing error is the error associated with activities conducted once survey responses have been received. It includes all data handling activities after collection and prior to estimation. Like all other errors, they can be random in nature, and inflate the variance of the estimates, or systematic, and introduce bias. At each data processing step a picture of the output files is compared to the ones from the previous step. This greatly improved the data processing stage.
Sampling error is defined as the error that results from estimating a population characteristic by measuring a portion of the population rather than the entire population. The most commonly used measure to quantify sampling error is sampling variance. The standard error of an estimator (the square root of its sampling variance) is easier to interpret since it provides an indication of sampling error using the same scale as the estimate whereas the variance is based on squared differences.
It is more useful in many situations to assess the size of the standard error relative to the estimate of the characteristic being measured. The coefficient of variation (CV) provides such a measure. It is the ratio of the standard error of the survey estimate to the average value of the estimate itself, across all possible samples. It is very useful in comparing the precision of sample estimates, where their sizes or scale differ from one another.