Labour force survey 2021
LFS 2021
LFS 2021
Consulter
Sommaire
Traitement statistique
Frequency of data collection
Continuous
Data collection
The survey is conducted continuously, every week of the year.
Data collection is carried out under computer-assisted collection (CAPI), face-to-face for the first interview of the household, and by telephone or internet for the re-interviews.
An annual specific survey is conducted in Mayotte.
Data collection period
The data collection is conducted continuously, every week of the year. The interview is conducted in the 3 weeks following the reference week.
Each dwelling is surveyed six consecutive quarters.
Collection mode
- Face to face by interviewer
- By Internet
- By phone
Survey unit
Household
Sampling method
- The sampling unit is dwelling. The sample is rotational: each cluster is surveyed 6 consecutive quarters, then replaced by a new cluster of the same sector.
- In Metropolitan France, the sampling frame is the demographic file of dwellings and individuals (<< Fideli >>). The sample selection is stratified by region. For each region, the sample is balanced, proportionnaly of the number of main residences, on some variables from the sampling frame.
- For DOM, the sampling is stratified (geographic strata) and systematic (with equal inclusion probabilities, within geographic strata).
- The sample is uniformely distributed among the 13 weeks of the quarter. Each cluster is affected to a reference week.
- The average sampling rate is 1/400.
- The complementary module is asked to the ingoing clusters.
Sample size
In 2021, each quarter, 80 000 dwellings are surveyed.
Data collection documents
- Questionnaire LFS 2021 (pdf, fr, 2 Mo, 16/11/2021)
Data compilation
- For metropolitan France, a two-steps weighting method is applied: calibration and non-response model on variables from the sampling frame in the first step, calibration on external data (population by age, sex and region; number of dwellings) in the second step.
- For DOM, a two-steps weighting method is also applied: non-response model on variables from the sampling frame in the first step, calibration on external data (population by age, sex and region; number of dwellings) in the second step.