Survey of Older Workers
Detailed information for 2008
The survey is designed to assess the labour market intentions and transitions of older Canadians.
Data release - December 2, 2009
- Questionnaire(s) and reporting guide(s)
- Data sources and methodology
- Data accuracy
The Survey of Older Workers (SOW) is sponsored by the Labour Market Policy branch of Human Resources and Skills Development Canada (HRSDC). The subject matter is intended to identify "factors" that influence the decision to retire or remain working. In this context pensions, general finances, the role of dependents, the nature of work, health considerations etc., will be of primary concern in trying to understand workers intentions and motivations.
These data support labour policy initiatives in the federal government. This informs policy development and may assist in future decision-making. Survey data will be used to:
- Provide national and provincial level estimates of households with workers aged 50 years and older and estimates on whether these workers experienced unemployment after the age of 50;
- Provide national level information on reasons for a spell of unemployment and what may follow a spell of unemployment;
- Provide national level estimates of the number of years working for their last employer;
- Provide national level estimates of the types of considerations that might encourage recent retirees to re-enter the labour force;
- Provide national level estimates on the main source of information in learning about retirement planning;
- Provide national level estimates of work/retirement plans by industry, and occupation;
- Provide national level estimates of work/retirement plans by type of pension plan;
- Provide national level estimates of identified barriers to saving for retirement.
- Work transitions and life stages
Data sources and methodology
The survey was administered to a sub-sample of the dwellings in the Labour Force Survey (LFS, record number 3701) sample, and therefore its sample design is closely tied to that of the LFS.
The LFS is a monthly household survey whose sample of individuals is representative of the civilian, non-institutionalized population 15 years of age or older in Canada's ten provinces. Specifically excluded from the survey's coverage are residents of the Yukon, Northwest Territories and Nunavut, persons living on Indian Reserves, full-time members of the Canadian Armed Forces and inmates of institutions. These groups together represent an exclusion of approximately 2% of the population aged 15 or over.
For the SOW, the coverage of the LFS was modified to include all members of the household aged 50 to 75 who were either still working or who had retired within the last 24 months.
The Statistics Canada Questionnaire Design Resource Centre (QDRC) was consulted. Three one-on-one test sessions and one focus group session was conducted in the development of the survey. In addition, the survey was presented to and approved by the Questionnaire Review Committee.
This is a sample survey with a cross-sectional design.
The Survey of Older Workers was administered from October to December 2008 to a sub-sample of the dwellings in the Labour Force Survey (LFS, record number 3701) sample, and therefore its sample design is closely tied to that of the LFS.
The SOW used five of the six rotation groups in the LFS sample. Rotation groups 2, 5 and 6 were selected in October and 1 and 3 in November. For the SOW, the coverage of the LFS was modified to include all members of the household aged 50 to 75 who were either still working or who had retired within the last 24 months.
Data collection for this reference period: 2008-10-19 to 2008-12-01
Responding to this survey is voluntary.
Data are collected directly from survey respondents.
Data were collected using computer-assisted telephone interviewing (CATI). The survey was non-proxy. It is an independent LFS supplement so that (if needed) call-backs were made to complete the survey once the LFS was done. The SOW was administered to one randomly selected individual per household. The random selection was carried out at the time of the interview.
Upon completion of the Labour Force Survey interview, the interviewer asked to speak to the selected person for the SOW. If the selected person was not available, the interviewer arranged for a convenient time to phone back. Proxy response was not allowed unless it was to aid in communicating with the selected respondent (i.e., translation, disability, etc.), and the selected respondent was present. Hence the collection period was extended by one week to allow the interviewers time to contact the pre-selected individuals.
View the Questionnaire(s) and reporting guide(s) .
Error detection was built into the CATI application (Blaise). Where applicable, these included consistency edits to resolve data disagreements, flow edits to be certain applicable questions where predictably populated and range edits to verify values beyond those expected.
There were no major problems encountered during processing. However, it is worthwhile to mention that 2,023 SOW records were removed at the clean-up step. It was found that these records either did not have sufficient information to be kept or responded to the LFS but refused to respond to the SOW.
Due to the large amount of editing built into the CATI application, the SOW dataset only required two edits. No other data items were imputed, deterministically or otherwise.
No imputation is done for this statistical program.
The principles behind the calculation of the weights for the SOW are identical to those for the LFS. However, further adjustments are made to the LFS sub-weights in order to derive a final weight for the individual records on the SOW microdata file.
1. Adjustment for computer-assisted personal interview (CAPI) and LFS non-response.
All LFS sample units appearing on the LFS Master file (TABS file) were assigned a survey weight. However, not all units on the TABS file were respondents in the month the data was collected. In some cases when a response could not be obtained for whatever reason, data for the month were imputed, often by carrying forward data from a previous month. Since these were actually cases of LFS non-response and since the SOW was to be selected from the LFS respondents, a separate non-response adjustment was made to account for these cases prior to the SOW weighting. The LFS non-response adjustment groups were used for this re-weighting step.
As well, LFS respondents for which data was collected by personal interview were excluded from the SOW sample. These sample units were grouped with LFS non-respondents and included in the above mentioned non-response adjustment.
2. Adjustment for rotation groups
The LFS sample is comprised of six rotation groups while the SOW sample uses only five of them. Therefore an adjustment of 1.2 is applied to the SOW weight to adjust for the missing rotation group.
3. Adjustment for sub-sampling
In selecting the sample for the SOW, all LFS households in the five SOW rotation groups with at least one person in the target population for the SOW were contacted for the survey. If there was more than one person in the household belonging to the target population, then one person was selected at random for the SOW sample. To account for sub-sampling, the SOW weight is multiplied by the number of household members belonging to the target population.
4. Adjustment for non-response
To adjust for non-response to the SOW, logistic regression was used to estimate the expected probability of response for each sample unit. Modeling was done within province. To form response groups for the adjustment, the sample file was sorted by probability of response and then divided into deciles giving ten response adjustment groups for each province. Within the adjustment groups, the weight adjustment factor was computed as the ratio of the sum of the weights for all sample units to the sum of the weights for all respondents.
5. Adjustment for calibration to projected population totals
A calibration of non-response adjusted weights is then performed. Non-response adjusted SOW weights for SOW respondents are combined with LFS sub-weights for LFS respondents aged 50 to 75 and not in the SOW target population. A calibration is then done adjusting weights so that the weights sum to known demographic estimates for the 50 to 75 year old age group. Calibration had to be done including the LFS non-SOW sample since the demographic estimates were only available for the whole population of 50 to 75 year olds and not for the SOW target population.
Considerable time and effort was made to reduce non-sampling errors in the survey. Quality assurance measures were implemented at each step of the data collection and processing cycle to monitor the quality of the data. These measures include extensive training of interviewers with respect to the survey procedures and computer-assisted telephone interviewing (CATI) application, observation of interviewers to detect problems of questionnaire design or misunderstanding of instructions and testing of the CATI application to ensure that range checks, edits and question flow were all programmed correctly.
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.
The following summarizes the assessment of risk that was done and the remedial measures taken to ensure the confidentiality of the file.
Very few sensitive variables appear on the file. All continuous variables (except for the survey weights) have been recoded into categorical variables.
The variable economic family size was capped at 3.
Regrouping of variables:
Extensive collapsing of identifying variables was done. With the level of detail remaining in the identifying variables, it is felt that there is little risk of identifying an individual spontaneously on the public use microdata file (PUMF).
Three way table analysis:
Following an analysis of three way tables, variable data was suppressed on the 1% of records with the highest multiplicities.
Risk factor analysis:
Risk factor analysis was performed to identify those records with the rarest combination of variable values among indirect identifiers. It was felt that the records identified did not pose a serious risk of disclosure.
Assessment of the risk of linking to the LFS PUMF:
The possibility of linking to the LFS PUMF was a concern. The rate at which records could be uniquely matched to the LFS dictated which LFS variables were ultimately included on the SOW PUMF. Of those that could be uniquely matched to the LFS PUMF, a high percent were correct matches. However, it was felt in the end, that because so little identifying information is coming from the SOW, the risk of disclosure through linkage was not great.
Review of weights:
A review of the SOW weights indicated that it was unlikely that sub-provincial geographic detail could be guessed at through examination of the weights. As well, since the initial weights, which came from the LFS, were subject to multiple adjustments to derive final weights, it was felt that it was unlikely that weights could be used to link the SOW units to their original LFS design strata.
Revisions and seasonal adjustment
This methodology does not apply to this survey.
While considerable effort is made to ensure high standards throughout all stages of collection and processing, the resulting estimates are inevitably subject to a certain degree of error. These errors can be broken down into two major types: non-sampling and sampling.
Non-response is an important source of non-sampling error.
The response rate for this survey was 79.5%.
Due to the high response rate for the survey, the coefficients of variation (CV) for key variables are very low and within Statistics Canada's pre-determined useable ranges most of the time. Respondents to the Displaced Workers (DW) section experienced a more detailed eligibility criteria within the application, so that section contains less respondents than previous sections of the questionnaire, resulting in higher CVs some of the time.
- Survey of Older Workers: Microdata User Guide 2008