Employment and unemployment (Labour force survey) (employ)

National Reference Metadata in ESS Standard for Quality Reports Structure (ESQRS)

Compiling agency: STATBEL


Eurostat metadata
Reference metadata
1. Contact
2. Statistical presentation
3. Statistical processing
4. Quality management
5. Relevance
6. Accuracy and reliability
7. Timeliness and punctuality
8. Coherence and comparability
9. Accessibility and clarity
10. Cost and Burden
11. Confidentiality
12. Comment
Related Metadata
Annexes (including footnotes)
 



For any question on data and metadata, please contact: Eurostat user support

Download


1. Contact Top
1.1. Contact organisation

STATBEL
1.2. Contact organisation unit

Thematic Division Society
1.5. Contact mail address

North Gate - Koning Albert II-laan 16 - 1000 Brussels


2. Statistical presentation Top
Please take note of the abbreviations used in the report 
Abbreviation Explanation
CV Coefficient of variation (or relative standard error)
Y/N Yes / No
H/P Households/Persons
M? Member State doesn’t know
NA Not applicable/ Not relevant
UNA Information unavailable
NR Non-response: Member State doesn’t answer to Eurostat request for information. Blank is allowed only in boxes with comments
LFS Labour Force Survey
NUTS Nomenclature of territorial units for statistics or corresponding statistical regions in the EFTA and candidates countries
2.1. Data description
Coverage   
Coverage Household concept Definition of household for the LFS Inclusion/exclusion criteria for members of the household Questions relating to employment status are put to all persons aged ...
 Private households Housekeeping  Members living together in the same dwelling, sharing meals and expenditures (criteria not explicitly checked in sampling frame and during interview) Households, all members of which are 90 or older, are excluded before the sample is drawn.  15-89

 

Population concept  Specific population subgroups
Primary/secondary students Tertiary students People working out of family home for an extended period for the purpose of work People working away from family home but returning for weekends Children alternating two places of residence
 Registered population  Family home  Family home  Family home  Family home  Family home

 

Reference week
Fixed week (data collection refers to one reference week, to which the observation unit has been assigned prior to the fieldwork) Rolling week (data collection always refers to the week before the interview)                                  
 X  

 

Participation is voluntary/compulsory?
 Compulsory
2.2. Classification system

[not requested for the LFS quality report]
2.3. Coverage - sector

[not requested for the LFS quality report]
2.4. Statistical concepts and definitions

[not requested for the LFS quality report]
2.5. Statistical unit

[not requested for the LFS quality report]
2.6. Statistical population

[not requested for the LFS quality report]
2.7. Reference area

[not requested for the LFS quality report]
2.8. Coverage - Time

[not requested for the LFS quality report]
2.9. Base period

[not requested for the LFS quality report]


3. Statistical processing Top
3.1. Source data
Sampling design & procedure

Sampling design (scheme; simple random sample, two stage stratified sample, etc.)

 Base used for the sample (sampling frame)  Last update of the sampling frame (continuously updated or date of the last update) Primary sampling unit (PSU)   Final sampling unit (FSU)
Two-stage stratified sampling. National Population Register (NPR), extended with information from tax and unemployment registers. For the primary sampling units, the last update was in September 2020. For the secondary (i.e. final) sampling units, the last update is made one or two months before the start of the quarter. Statistical section, or part of a sub-municipality consisting of several statistical sections.

We select the private household at the dwelling. In case that another household is living in the dwelling, this household will not be interviewed.  

 

Sampling design & procedure
First (and intermediate) stage sampling method   Final stage sampling method Stratification (variable used) Number of strata (if strata change quarterly, refer to Q4). Rotation scheme (2-2-2, 5, 6, etc..)

The first stage sampling frame, i.e. the frame, consists of geographic areas, which are either 'statistical sections' or unions of statistical sections within 'statistical letters' or sub-municipalities. Preferably neighbouring sections are joined together, but this is not the rule.  Each PSU must contain a minimum number of eligible private households: if a PSU is sampled in the first stage, it must be possible to select enough households (26 in Brussels Capital Region, and 23 in other sampling strata) to form a 'group of households' (which will be assigned as a whole to an interviewer). There are about 6354 PSUs in the sampling frame, containing 676 households on average; ‘small’ sections only represent 0.15% of the total number of households. The frame of PSUs is stratified by and sorted on some PSU characteristics (for details, see the box headed “Stratification” to the right in this table).

Systematic probability proportional to size sampling (PPS-SYS; where a proxy for the number of eligible private households is used as size measurement) is applied in each first stage sampling stratum. The number of PSU draws or selections in each sampling stratum is fixed in advance, such that 6695 HHs are selected in total per quarter in the second stage. Larger PSUs can be selected more than once, while smaller PSUs are probably not selected.

Important to notice is that each PSU draw, i.e. each 'group of households', is assigned to a reference week immediately after the first stage. The 286 groups of households that are selected in total each quarter, are uniformly spread over the 13 weeks of the quarter (i.e. 22 groups per week).

In the second, and final, stage of sampling, the PSUs sampled in stage 1 technically serve as sampling strata.

The number of households selected in stage 2 (the final sampling stage) in a selected PSU equals the number of times the PSU is selected in the first stage times the size of the groups of households to be formed in the PSU (i.e. 26 in the Brussels Capital Region and 23 in the other strata). Basically, simple random sampling (SRS) is applied to select households (SSUs or FSUs) in each selected PSU.

Stratification of private households in the second stage is applied since the first quarter of 2021. Two strata are considered in this stage: households containing at least one 15-74 year old person – called type 1 households –, versus households containing no 15-74 year old, but at least one 75-89 year old person – called type 2 households. If a PSU is selected only once in the first sampling stage, STR-SRS is applied to a frame of eligible households, where eligibility means that a household is of type 1 or type 2 considering household members’ age at the reference Sunday, i.e., the last day of the reference week, which is determined after the first sampling stage. If a PSU is selected more than once, the “groups” of households are selected (by STR-SRS) one after another, where the frame of eligible households is adapted to the specific group (i.e., to its reference week/Sunday), considering that households should be selected for only one group. Each selected “group” of households contains exactly 1 type 2 household; the other 25 (=26–1, in the Brussels-Capital Region) or 22 (=23–1, in the other PSU-strata) are type 1 households.

The new sample of households/individuals selected in each quarter, is called a rotation group (RG).

In the first stage of sampling, the PSUs are explicitly stratified by NUTS 2 regions (i.e. provinces, where the Brussels Capital Region and the German community are separate strata). The PSU sampling frame, within each stratum, is further sorted on (1) the quintile of the number of private households in the PSU, (2) the quintile of the unemployment rate in the PSU and (3) the quintile of the average household income in the PSU. This implies implicit stratification on the latter three PSU characteristics within each explicit stratum. Serpentine sorting is applied.

Stratification of private households is applied at sampling stage 2. Two strata are considered in this stage: households containing at least one 15-74 year old person – called type 1 households –, versus households containing no 15-74 year old, but at least one 75-89 year old person – called type 2 households. 

12 (explicit) strata at stage 1 in each quarter; the stratification variable is denoted PROV12.

2 strata at stage 2; the stratification variable is denoted HHstrat.

 2-(2)-2

 

Yearly sample size & Sampling rate
Overall theoretical yearly sampling rate Size of the theoretical yearly sample
(i.e. including non-response) (i.e. including non-response)
 0,548%  107120 households

  

Quarterly sample size & Sampling rate

Overall theoretical quarterly sampling rate

Size of the theoretical quarterly sample

(i.e. including non-response)

(i.e. including non-response)
0.137%  26780 households

 

Use of subsamples to survey structural variables (wave approach)

Only for countries using a subsample for yearly variables
 Wave(s) for the subsample  Are the 30 totals for ILO labour status (employment, unemployment and inactivity) by sex (males and females) and age groups (15-24, 25-34, 35-44, 45-54, 55+) between the annual average of quarterly estimates and the yearly estimates from the subsample all consistent? (Ref.: Commission Reg. 430/2005, Annex I) (Y/N) If not please list deviations List of yearly variables for which the wave approach is used (Ref.: Commission Reg. 377/2008, Annex II)
Wave 1 Y  NA All except 'HOMEWK', 'TEMPREAS', 'TEMPAGCY' . These variables are collected in all waves.

 

Brief description of the method of calculating the quarterly core weights Is the sample population in private households expanded to the reference population in private households? (Y/N) If No, please explain which population is used as reference population Gender is used in weighting (Y/N) Which age groups are used in the weighting (e.g., 0-14, 15-19, ..., 70-74, 75+)? Which regional breakdown is used in the weighting (e.g. NUTS 3)? Other weighting dimensions

The 2-step quarterly weighting model can be formulated as: < IND; STRAT12 × SEX × AGECAT + RG_c; d⁄p ; Lin >.

The third component,  d⁄p, in this formal representation of the weighting model, means that, in step 1, the sampling weights d are corrected using estimated response probabilities p at household level; a random intercept logistic regression model, followed by smoothing, is used to estimate the p.

In step 2, proper calibration is applied to further adjust the corrected weights d/p. The first component, IND, within < > indicates that calibration is at individual level. The second component is a formal expression for the linear structure of the calibration model in step 2, indicating that (1) calibration is to the joint distribution of variables STRAT12, SEX and AGECAT in the population, and (2) the totals of calibrated weights for the RGs involved in each quarterly sample are forced to be proportional to the initial sizes of these RGs (the notation RG_c stands for “contrast constraints between RGs”).  The fourth and last component, Lin, indicates the use of the linear method for calibration.		  Y NA Y, i.e. through the variable SEX as mentioned to the left. Variable AGECAT identifies fifteen 5-year age classes (0-4, 5-9, …, 70-74) and the open ended sixteenth age class 75+. Variable STRAT12, which is also the stratification variable in the first sampling stage, corresponds to NUTS 2 level.

The random intercept logistic regression model used to estimate the households’ response probabilities includes fixed effects variables household type, household origin, STRAT12 and level of urbanisation, and the random effects variable PSU identification. This regression model is applied to each RG separately, for each trimester.

 

 

Brief description of the method of calculating the yearly weights (please indicate if subsampling is applyed to survey yearly variables) Gender is used in weighting (Y/N) Which age groups are used in the weighting (e.g., 0-14, 15-19, ..., 70-74, 75+)? Which regional breakdown is used in the weighting (e.g. NUTS 3)? Other weighting dimensions

 The 2-step weighting model can be formulated as: < IND; STRAT12 × SEX × AGECAT + REGION × SEX × AGECAT* × StatBIT; d⁄p; Lin >. This model is comparable to the quarterly weighting model already explained. The differences are: (1)  application of this model to the wave 1 sub-sample of respondents; (2) calibration to not only the joint distribution of STRAT12, SEX and AGECAT in the population, but also to the joint distribution of REGION, SEX, AGECAT* and StatBIT, estimated as full-sample annual averages from LFS; (3) there are no between-RG balancing (or “contrast”) constraints. The new term REGION × SEX × AGECAT* x StatBIT in the linear structure of the model allows to satisfy the consistency requirements imposed by Eurostat.

Exceptionally, consistency on the level of region has to be dropped because of model convergence issues. In 2021, this was the case for the fourth quarter. Nevertheless, the minimum consistency requirements imposed by Eurostat were still met

  Y, i.e. through the variable SEX as mentioned to the left. Variable AGECAT (16 classes) is defined as before; variable AGECAT* identifies 7 age classes 0-15, 15-24, 25-34, 35-44, 45-54, 55-64 and 65+.  Variable STRAT12 (NUTS 2 level) is as before; variable REGION is NUTS 1 level. The regression model to estimate the response probabilities is exactly as for the quarterly weighting model.

 

Brief description of the method of calculating the weights for households External reference for number of households etc.? Which factors at household level are used in the weighting (number of households, household size, household composition, etc.) Which factors at individual level are used in the weighting (gender, age, regional breakdown etc.) Identical household weights for all household members? (Y/N)
NA  NA NA  NA  NA 
3.2. Frequency of data collection

[not requested for the LFS quality report]
3.3. Data collection
Data collection methods: brief description Use of dependent interviewing (Y/N)? In case of Computer Assisted Methods adoption for data collection, could you please indicate which software is used?

For the first wave the detailed information (related to individuals aged 15 years and over) is collected by means of face-to-face interviews in the 3 or 4 weeks following the week of reference. In households of retired persons, interviews can be conducted by telephone. Exceptionally, in 2021, due to the Covid-crisis, all interviews in the first wave were done in CATI.


For the 2nd to 4th wave, the data are collected using Web (CAWI) or telephone (CATI). The household itself decides in which mode they want to take the next survey. In case of nonresponse in CAWI, the interviewer tries to convince the household to participate and then the interview might take place in CATI. During the second and following waves, interviewers try to reach only the respondents of the previous wave. The previous non-respondents (for wave n) stay non-respondents (for wave n+1).

In case the selected household does not exist anymore at the moment of the survey in the field, we try to find the new address of the household and in case it is in the same municipality, the household can be interviewed there. The interviewer can also interview the household that now lives at the address but not both (the ‘old’ and ‘new’ household). 

  Y (but limited)

 (CAPI)/CATI/CAWI

Blaise software is used.

 

Are any LFS data collected from registers (Y/N)? If Yes, please indicate which variables are collected from registers.
 Y  INCGROSS
3.4. Data validation

[not requested for the LFS quality report]
3.5. Data compilation

[not requested for the LFS quality report]
3.6. Adjustment

[not requested for the LFS quality report]


4. Quality management Top
4.1. Quality assurance

[not requested for the LFS quality report]
4.2. Quality management - assessment

[not requested for the LFS quality report]


5. Relevance Top
5.1. Relevance - User Needs
Description of users with respect to the statistical data

Data are used by a diversity of users: policy makers, other stakeholders, media, academic research, students, general public...

There is a large demand for the Belgian LFS data and the data are used a lot in official studies and reports, for decision making and for scientific research.
Indication of the needs and uses for which users want the statistical outputs; information on unmet user needs and any plans to satisfy them in the future

Since the Covid crisis there is a hugh demand for monthly data so since Spring 2020 we publish experimental monthly figures on our website but we stress that they are experimental and produced for the specific purpose of monitoring the coronavirus crisis. The figures are provisional figures, produced based on a first version of the data and where the speed takes precedence over the completeness and quality of the data received. The figures can be found here: https://statbel.fgov.be/en/themes/datalab/monthly-figures-labour-market#figures

Even after the crisis users want to continue using the monthly figures.

 
5.2. Relevance - User Satisfaction

[not requested for the LFS quality report]
5.3. Completeness
NUTS level of detail   
Regional level of an individual record (person) in the national data set Lowest regional level of the results published by NSI Lowest regional level of the results delivered to researchers by NSI Brief description of the method which is used to produce NUTS-3 unemployment and labour force data sent to Eurostat?
 Municipality level NUTS 1 (region (gewest / région)) + a limited number of variables on NUTS 2 level.  NUTS 2 (province (provincie /province)) and in a limited cases (for specific uses) NUTS 3 or municipal data.  We do not produce NUTS-3 unemployment and labour force data.
5.3.1. Data completeness - rate

[not requested for the LFS quality report]


6. Accuracy and reliability Top
6.1. Accuracy - overall

[not requested for the LFS quality report]
6.2. Sampling error
Publication thresholds   
Annual average estimates Yearly estimates - wave approach 
 Limit below which figures cannot be published  Limit below which figures must be published with warning  Limit below which figures cannot be published Limit below which figures must be published with warning
 2000  5000  3000  8000
Biennial variables estimates Household estimates Household average estimates
Limit below which figures cannot be published Limit below which figures must be published with warning Limit below which figures cannot be published Limit below which figures must be published with warning Limit below which figures cannot be published Limit below which figures must be published with warning
 3000  8000  3000  8000  3000  8000
6.2.1. Sampling error - indicators
Coefficient of variation (CV) Annual estimates
Sampling error - indicators - Coefficient of variation (CV), Standard Error (SE) and Confidence Interval (CI)       
 

            Employment rate                                 

Unemployment-to-population ratio                      

Youth unemployment rate as a percentage of labour force
 

Age group: 15 -74

Age group: 15 -74

Age group: 15 -24
 CV  0.35%  2.24%  4.65%
 SE  0.20% 0.08%  0.85% 
 CI(**)  56.42% - 57.19%  3.64% - 3.97%  16.55% - 19.87% 
                                      Unemployment-to-population ratio 15-74 (NUTS 2 regions)                                 
  CV       SE         CI(**)  
BE10 Brussel  4.44% 0.32%   6.82% - 8.07%
BE21 Antwerpen  7.49% 0.24%  2.76% - 3.70% 
BE22 Limburg  9.16% 0.18%  1.72% - 2.42% 
BE23 Oost-Vlaanderen  9.80% 0.18%  1.54% - 2.23% 
BE24 Vlaams-Brabant  7.21% 0.18%  2.12% - 2.82% 
BE25 West-Vlaanderen  7.58% 0.17%  1.96% - 2.63% 
BE31 Waals-Brabant  8.45% 0.36%  3.44% - 4.85% 
BE32 Henegouwen  5.75% 0.32%  4.91% - 6.15% 
BE33 Luik  5.89% 0.31%  4.75% - 5.97% 
BE34 Luxemburg  8.38% 0.29%  2.93% - 4.07% 
BE35 Namen  9.64% 0.41%  3.56% - 5.15% 

 

Description of the assumption underlying the denominator for the calculation of the CV for the employment rate
 The denominator of the employment rate is treated as a population figure without sample variance

 

Reference on software used: Reference on method of estimation:
 Standard error is estimated using the SAS POULPE macro, taking into account sampling design, non-respons correction, calibration procedure and linearization of ratios.  NA

  

(*) The coefficient of variation for actual hours worked should be calculated for the sum of actual hours worked in 1st and 2nd jobs, and restricted to those who actually worked 1 hour or more in the reference week.

(**) The value is based on a CI of 95%. For the rates the CI should be given with 2 decimals.
6.3. Non-sampling error

 [not requested for the LFS quality report]
6.3.1. Coverage error
Frame quality (under-coverage, over-coverage and misclassifications(b))      
Under-coverage rate (%) Over-coverage rate (%) Misclassification rate (%)  Comments: specification and impact on estimates(a)   
 Undercoverage  Overcoverage  Misclassification(b)  Reference on frame errors
 <0.5%  very small   very small Households, all members of which are 90 years or older, and collective households (about 0,15% of all households) are excluded before draw. Delay between draw of household (from National Population Register, kept up to date "permanently") and fieldwork: between 2 and 6 months  M? M?   UNA

 


(a) Mention specifically which regions / population groups are not suitably represented in the sample.
(b) Misclassification refers to statistical units having an erroneous classification where both the wrong and the correct one are within the target population
.
6.3.1.1. Over-coverage - rate

[Over-coverage rate, please see concept 6.3.1 Coverage error in the LFS quality report]
6.3.1.2. Common units - proportion

[not requested for the LFS quality report]
6.3.2. Measurement error
Errors due to the medium (questionnaire)   
Was the questionnaire updated for the 2020 LFS operation? (Y/N) Synthetic description of the update Was the questionnaire tested? (Y/N) If the questionnaire has been tested, which kind of tests has been applied (pilot, cognitive, internal check)?
 Y, IESS revision in 2021  Whole questionnaire has been revised  Y Internal testings + Statistics Belgium ran a two-wave pilot survey during all four quarters of 2020 (except for Q1: only the 1st wave), for a sample size approximately ¼ of the size of the regular first and second waves of the survey. 

 

Main methods of reducing measurement errors 
Error source  
Respondent  Letter introducing the survey (Y/N) Phone call for booking or introducing the survey (Y/N)
 Y  Y (but not always)
Interviewer  Periodical training (at least 1 time per year) (Y/N)  Feedbacks from interviewer (reports, debriefings, etc.) (Y/N)
 Y  Y
Fieldwork  Monitoring directly by contacting the respondents after the fieldwork (Y/N) Monitoring directly by listening the interviews (Y/N) Monitoring remotely through performance indicators (Y/N)
Y   N  N
Questionnaire  Questionnaire in several languages (Y/N)  On-line checks (for computer assisted interviews (Y/N)
Y   Y
Other / Comments  No
6.3.3. Non response error

[not requested for the LFS quality report]
6.3.3.1. Unit non-response - rate

IN THIS SECTION INFORMATION REFERS TO THE FINAL SAMPLING UNITS *

Methods used for adjustments for statistical unit non-response
Adjustment via weights (Y/N) Variables used for non-response adjustment Description of method
 Y NUTS 2, degree of urbanisation, household type and origin of household Adjustment for unit (household) non-response is realized as a correction of the sampling weights: each responding household's sampling weight is divided by an estimate of its response probability, providing '(non-)response adjusted weights' that are further adjusted by calibration.　

Response probabilities (at household level) are estimated through a 'random intercept logistic regression model'. The fixed effects included are main effects (i.e. no interaction effects are included) from four variables: (1) household type, (2) household origin, (3) Nuts II regions (i.e. stage 1 sampling strata) and (4) degree of urbanisation. The intercept further includes random effects from PSUs. The resulting estimated response probabilities are smoothed to avoid extreme and too volatile values: the random part of the linear predictor is grouped into quintiles, and the values in each quintile are averaged. This smoothing technique does not significantly reduce the predictive power of the model.
Substitution of non-responding units (Y/N) Substitution rate Criteria for substitution
 N  NA  NA
Other methods (Y/N) Description of method
 N  In case that households refuse to participate or cannot be contacted, we do not try to re-contact them in the following waves. The data below in wave 2 and following are thus new refusals or non-contacts. Migration is only registered in wave 1. 

  

ATTENTION: NON-RESPONSE RATE IS THE RATIO OF THE NUMBER OF NON RESPONDING UNITS TO THE TOTAL NUMBER OF THE ELIGIBLE UNITS IN THE SAMPLE

Non-response rates by survey mode. Annual average (% of the theoretical yearly sample by survey mode)
Survey
CAPI CATI  PAPI  CAWI  POSTAL
NA UNA  NA  UNA  NA

 

Divisions of non-response into categories. Quarterly data and annual average  (% of the theoretical quarterly sample)
Quarter Non-response rate
Total (%)             of which:
 Refusals (%)      Non-contacts (including people who migrated (or moved) internally or abroad) (%)     of which people who migrated (or moved) internally or abroad (%)
1 18.5% 3.7%  5.4%  UNA
2  22.9% 4.3%  8.4%  UNA 
3  21.0% 3.5%  7.9%  UNA 
4  21.3% 4.3%  7.8%  UNA 
Annual

20.9%

 4.0% 7.4% UNA 

 


 

 Units who refused to participate in the survey  (Please indicate the number of the units concerned in the cells where the wave is mentioned)
Subsample
Quarter1_y Quarter2_y Quarter3_y Quarter4_y
Subsample_Q4_Y-2

wave 4: 40

      
Subsample_Q1_Y-1 wave 3: 44 wave 4:28    
Subsample_Q2_Y-1   wave 3:55 wave 4:23  
Subsample_Q3_Y-1     wave 3:44 wave 4:25
Subsample_Q4_Y-1 wave 2: 45     wave 3:32
Subsample_Q1_Y wave 1: 564 wave 2:84    
Subsample_Q2_Y   wave 1:662 wave 2:74  
Subsample_Q3_Y     wave 1:479 wave 2:57

Subsample_Q4_Y

       wave 1:667
Total in absolute numbers        
Total in % of theoretical quarterly sample        


 

 Units who were not contacted (including people who migrated (or moved) internally or abroad) (Please indicate the number of units only in the cells where the wave is mentioned)
Subsample
Quarter1_y Quarter2_y Quarter3_y Quarter4_y
Subsample_Q4_y-2 wave 4:81      
Subsample_Q1_y-1 wave 3:136 wave 4:107    
Subsample_Q2_y-1   wave 3:191 wave 4:114  
Subsample_Q3_y-1     wave 3:168 wave 4:98
Subsample_Q4_y-1 wave 2:126     wave 3:156
Subsample_Q1_y wave 1:687 wave 2:235    
Subsample_Q2_y   wave 1:1003 wave 2:161  
Subsample_Q3_y     wave 1:938 wave 2:218
Subsample_Q4_y       wave 1:925
Total in absolute numbers        
Total in % of theoretical quarterly sample        

 

 of which people who migrated (or moved) internally or abroad) (Please indicate the number of units only in the cells where the wave is mentioned)
Subsample
Quarter1_y Quarter2_y Quarter3_y Quarter4_y
Subsample_Q4_y-2 wave 4      
Subsample_Q1_y-1 wave 3 wave 4    
Subsample_Q2_y-1   wave 3 wave 4  
Subsample_Q3_y-1     wave 3 wave 4
Subsample_Q4_y-1 wave 2     wave 3
Subsample_Q1_y wave 1:81 wave 2    
Subsample_Q2_y   wave 1:74 wave 2  
Subsample_Q3_y     wave 1:107 wave 2
Subsample_Q4_y       wave 1:79
Total in absolute numbers        
Total in % of theoretical quarterly sample        

 

Non-response rates. Annual average (% of the theoretical yearly sample)
NUTS-2 region (code + name)  Non response rate (%)
BE21-Antwerpen  21.2
BE10-Bruxelles / Brussel  22.9
BE22-Limburg  21.0
BE23-Oost-Vlaanderen  23.7
BE24-Vlaams-Brabant  16.9
BE25-West-Vlaanderen  14.3
BE32-Brabant wallon  26.0 
BE32-Hainaut  23.1 
BE33-Liège  25.7 
BE34-Luxembourg  13.3
BE35-Namur 18.7

 

* If the final sampling unit is the household it must be considered as responding unit even in case of some household members (not all) do not answer the interview

 
6.3.3.2. Item non-response - rate
Item non-response - Quarterly data (Compared to the variables defined by the Commission Implementing Regulation (EC) No 2019/2240)       

Variable status

 Column Identifier Quarter 1 Quarter 2 Quarter 3 Quarter 4 Short comments on reasons for non-available statistics and prospects for future solutions

Compulsory / optional
 Compulsory  269-270  LEAVREAS  34.6% 32.4%   31.4%  16.1%  

We only started requesting LEAVREAS on a quarterly basis from 2021. That's the reason why in 2021 we could only take over a value from a previous wave when:

- the rotation group started in 2021

- the rotation group was in a second wave in 2021 (in that case the value from wave 1 could be taken, even if wave 1 fell in 2020)
 Compulsory  210-212 REGIONW          Similar to LEAVREAS, we only started collecting NUTS 2 region for people working in one of the border regions of neighbouring countries from 2021 so we could only gradually start to take over the info from the previous wave. In case the info is taken from a wave in 2020, REGIONW is not missing but it contains the country code instead of the NUTS 2 region code.
Compulsory 171 COBFATH 19.6% 19.4% 19.9% 20.3% The item non-response on COBFATH and COBMOTH is due to missing information in the National Population Register, which is currently the only source used for these variables. Actions will be undertaken to complement the missing information with new questions in the questionnaire from 2023 on. 
Compulsory 174 COBMOTH 17,0% 16,8% 17,3% 17,6% The item non-response on COBFATH and COBMOTH is due to missing information in the National Population Register, which is currently the only source used for these variables. Actions will be undertaken to complement the missing information with new questions in the questionnaire from 2023 on. 

 

Item non-response - Annual data (Compared to the variables defined by the Commission Implementing Regulation (EC) No 2019/2240)    
Variable status Column Identifier This reference year Short comments on reasons for non-available statistics and prospects for future solutions

 

 

 

 
6.3.4. Processing error

 

Editing of statistical item non-response

Do you apply some data editing procedure to detect and correct errors? (Y/N)

 Overall editing rate (Observations with at least one item changed / Total Observations )
 Y  M? (low)
6.3.4.1. Imputation - rate

 

Imputation of statistical item non-response
Are all or part of the variables with item non response imputed? (Y/N) Overall imputation rate (Observations with at least one item imputed / Total Observations )
 Y, Part of the variables  
 Main variables Imputation rate  Describe method used, mentioning which auxiliary information or stratification is used 
Level of education  very low Open responses are recoded.

If no response on open question, imputations for the highest level of education are based on the combination of sex, age classe, professional status and profession for employed persons. For unemployed or inactive persons, we take into account sex and age class.
REGIONW (place of work) very low Imputations are based on the place of work of previous job or the place of residence
NACE of the local unit very low Open responses are recoded.

If no response on open question, imputations for NACE take into account sex, region, professional status and profession.
ISCO very low

Imputation for isco takes into account economic activity, professional status, sex and age class.
SEEKDUR  15%

 Imputations are based on similar cases taking into account YEARPR and MONTHPR for respondents who had ever worked.
SIZEFIRM 0,2%

 First we tried to look up the sizefirm via the name and the municipality code of the firm. If this was not possible we looked at the sizefirm for the same detailed NACE code.
FINDMETH  37%

There was a program error in Q1 and Q2 and FINDMETH was only collected for respondents having found their job during the last 12 months instead of the last 7 years. A hot deck imputation was done using the information on type of contract, professional status, ISCO, NACE, age, sex, hatlevel. 
MIGREAS  15%

Because of a program error in Q1 and Q2, we could not distinguish between category 1 and 2 of MIGREAS. A hot deck imputation was done using information on country of birth, labour status, age, sex, hatlevel and ISCO. 

 

 
6.3.5. Model assumption error

[not requested for the LFS quality report]
6.4. Seasonal adjustment
Do you apply any seasonal adjustment to the LFS Series? (Y/N) If Yes, is your adopted methodology compliant with the ESS guidelines on seasonal adjustment? (ref. http://ec.europa.eu/eurostat/web/research-methodology/seasonal-adjustment) (Y/N) If Yes, are you compliant with the Eurostat/ECB recommendation on Jdemetra+ as software for conducting seasonal adjustment of official statistics. (ref. http://ec.europa.eu/eurostat/web/ess/-/jdemetra-officially-recommended-as-software-for-the-seasonal-adjustment-of-official-statistics) (Y/N) If Not, please provide a description of the used methods and tools
 N NA   NA  NA
6.5. Data revision - policy
Do you adopt a general data revision policy fully compliant with the ESS Code of Practice principles? (in particular see the 8th principle) (Y/N) Are you compliant with the ESS guidelines on revision policy for PEEIs? (ref. http://ec.europa.eu/eurostat/documents/3859598/5935517/KS-RA-13-016-EN.PDF) (Y/N)

 Y

Y (https://statbel.fgov.be/en/about-statbel/quality/revision-policy)
6.6. Data revision - practice

[not requested for the LFS quality report]
6.6.1. Data revision - average size

[not requested for the LFS quality report]


7. Timeliness and punctuality Top
7.1. Timeliness
Restricted from publication
7.1.1. Time lag - first result
Restricted from publication
7.1.2. Time lag - final result
Restricted from publication
7.2. Punctuality

All datasets were delivered on time.
7.2.1. Punctuality - delivery and publication

[not requested for the LFS quality report]


8. Coherence and comparability Top
8.1. Comparability - geographical

Divergence of national concepts from European concepts

(European concept or National proxy concept used) List all concepts where any divergences can be found

    
Is there a divergence between the national and European concepts for the following characteristics? (Y/N) Give a description of difference and provide an assessment of the impact of the divergence on the statistics
Definition of resident population (*)  Y  We use the Registered population definition in the sampling frame. At the start of the interview, the interviewer verifies with the respondents whether the list of household members as taken from the sampling frame corresponds to the actual situation.

Interviewers are instructed to drop household members that no longer live at the given address since at least 6 months (but in practice they sometimes drop household members earlier). Interviewers can also add new members in the household. Some of the most frequently added kind of persons are newborns, a new partner (or a partner from another member of the household), children in co-parenting that are registered at the address of the other parent, needy family members,...
Identification of the main job (*) N   
Employment N   
Unemployment N  In Belgium, unemployment rate is calculated and published for persons aged 15-64 and not 15 + or 15-74 (Eurostat)
8.1.1. Asymmetry for mirror flow statistics - coefficient

[not requested for the LFS quality report]
8.2. Comparability - over time
Changes at CONCEPT level introduced during the reference year and affecting comparability with previous reference periods (including breaks in series)
Changes in (Y/N) Description of the impact of the changes Statistics also revised backwards (if Y: year / N) Variables affected Break in series to be flagged (if Y: year and quarter/N)  
concepts and definition Y IESS; changes to the definitions of employment and unemployment Y (2009-2020) ILO status Y Q1 2021
coverage (i.e. target population) Y Population aged 15-89 instead of population aged 15+ N All  N
legislation Y  IESS NA NA  NA
classifications N NA  NA  NA  NA 
geographical boundaries N  NA  NA  NA  NA 

 

Changes at MEASUREMENT level introduced during the reference year and affecting comparability with previous reference periods (including breaks in series)
Changes to (Y/N) Description of the impact of the changes Statistics also revised backwards (if Y: year / N) Variables affected Break in series to be flagged (if Y: year and quarter/N)
sampling frame N  NA  NA  NA  NA
sample design Y   Before 2021, households, all members of which are 77 years or older were excluded before draw. Since 2021, households, all members of which are 90 years or older are excluded. Stratification of private households in the second stage is applied. Two strata are considered in this stage: households containing at least one 15-74 year old person – called type 1 households –, versus households containing no 15-74 year old, but at least one 75-89 year old person – called type 2 households.  N  N  N
rotation pattern N  NA   NA  NA  NA
questionnaire Y   IESS  N  All Y Q1 2021
instruction to interviewers Y  IESS  NA Most important: variables on ILO-status and working time and absences.  NA
survey mode N/Y  As part of the measures to limit the spread of the coronavirus, all face-to-face interviews have been temporarily replaced by telephone interviews since the first lockdown in March 2020 and during the whole year 2021. So all first wave interviews were CATI interviews in 2021.  The follow-up surveys are conducted - as before the Covid-19 crisis - by telephone or via the internet
  NA  NA  NA
weighting scheme N  NA  NA  NA  NA
use of auxiliary information N   NA  NA  NA  NA
8.2.1. Length of comparable time series

[not requested for the LFS quality report]
8.3. Coherence - cross domain
Coherence of LFS data with Business statistics data    
  Description of difference in concept Description of difference in measurement Give an assessment of the effects of the differences Give references to description of differences
Total employment

The total employment for workers in business statistics is calculated based on the number of people working in the enterprise in a limited number of sectors. In LFS it is about the employment of people living in the country in all sectors. 

The total number of self-dependent people as their main profession can not yet be determined in business statistics. 

 Timing, reference period, ...

 

Since 2021, people who are temporarily unemployed (on lay-ff) on a full-time basis for more than three months, are no longer counted as employed in LFS whereas they stay employed in business statistics.

Big impact (40,000 persons in 2021) of the lay-offs for more than three months. 

Other differences are difficult to assess.

  UNA
Total employment by NACE The NACE in business statistics has been assigned based on the added value and monitoring by NACE-experts while LFS NACE is an assesment of the activity of the respondent working in the local unit that is recoded by NACE-experts.  NACE in LFS is at the local unit, NACE in most business statistics are (currently) at the legal unit level Internal analyses have shown that there is some discrepancy, but this is overall rather limited. Most differences arise due to the limited description of the LFS respondents.   internal notes
Number of hours worked The number of hours worked in business statistics are the number of hours that are paid to the worker. 
  NA  These data can not be compared  UNA

 

Coherence of LFS data with registered unemployment  
Description of difference in concept Description of difference in measurement Give references to description of differences
ILO concept (LFS) versus administrative concept. There exist different administrative concepts. Here we take into account the persons without work, seeking work and receiving unemployment benefits (concept 'niet -werkende WZ UVW'). These statistics are published by the national employment office. Receiving unemployment benefits is not a condition for being ILO-unemployed. Survey versus administrative data. The national employment office publishes figures about persons receiving unemployment benefits which is not a condition for being ILO-unemployed.  UNA

 

Assessment of the effect of differences of LFS unemployment and registered unemployment     
Give an assessment of the effects of the differences          
Overall effect Men under 25 years Men 25 years and over Women under 25 years Women 25 years and over Regional distribution (NUTS-3)

The number of registred unemployed persons who are looking for a job (werkzoekende uitkeringsgerechtigde volledig werklozen, UVW-WZ, 321,502 is almost equal to the number of ILO unemployed persons (324,376) but there is a difference in concept and measurement.

Number of registered unemployed persons seeking work (concept niet werkende werkzoekenden) (464,070) much higher than the number of ILO unemployed persons because of a difference in measurement.

 

 more ILO unemployed men under 25  years than registered unemployed men (concept 'niet-werkende WZ UVW') more registered unemployed persons than ILO unemployed persons more ILO unemployed women under 25 years than registered unemployed men (concept 'niet-werkende WZ UVW') more registered unemployed persons than ILO unemployed persons  UNA
8.4. Coherence - sub annual and annual statistics

[not requested for the LFS quality report]
8.5. Coherence - National Accounts
Coherence of LFS data with National Accounts data    
  Description of difference in concept Description of difference in measurement Give an assessment of the effects of the differences Give references to description of differences
Total employment administrative  concept (national  concept) versus ILO concept 

NA results are based on administrative data (data of social security institutions). There are some additions/corrections f.i. for domestic staff, interim work, seamen, students work, undeclared work, … The employment figures based on LFS are without additions or corrections. LFS measures employment on a continuous basis and results can be seen as an average for a certain quarter or year, NA data measure employment on a certain date (end of the quarter).

Since 2021, people who are temporarily unemployed (on lay-ff) on a full-time basis for more than three months, are no longer counted as employed in LFS whereas they stay employed in NA.

higher number of employed persons in national accounts (5,104,200 in NA (national employment concept); 
4,853,682 in LFS)

Impact of the lay-offs for more than three months : 40,000 persons in 2021.

  internal notes
Total employment by NACE administrative concept (national concept) versus ILO concept NA: administrative data (declaration of the enterprise) LFS: declaration of the respondents + codification by experts. In NA interims are coded in nace2008 78, in LFS in the sector where they were working. NA data by NACE are of better quality compared to LFS figures UNA
Number of hours worked administrative concept (national concept) versus ILO concept  

NA: based on administrative data about hours paid + corrections. For hours worked of self employed use is made of LFS data. LFS: hours actually worked based on continuous survey (LFS)

  UNA  UNA

 

Which is the use of LFS data for National Account Data?   
Country uses LFS as the only source for employment in national accounts. Country uses mainly LFS, but replacing it in a few industries (or labour status), on a case-by-case basis Country not make use of LFS, or makes minimal use of it Country combines sources for labour supply and demand giving precedence to labour supply sources (i.e. LFS) Country combines sources for labour supply and demand not giving precedence to any labour side Country combines sources for labour supply and demand giving precedence to labour demand sources (i.e. employment registers and/or enterprise surveys)
 N N  Y (LFS is used in the calculation of hours worked of self employed persons)  N  N  N
8.6. Coherence - internal

[not requested for the LFS quality report]


9. Accessibility and clarity Top
9.1. Dissemination format - News release

[not requested for the LFS quality report]
9.2. Dissemination format - Publications
Please provide a list of type and frequency of publications

Every quarter different excel files with main results, flow statistics and our dynamic application beSTAT are updated on our website. The updates are associated with a press release about the main results and another about the flow statistics. When publishing the results of the first quarter of 2021 we put a note on the website about the changes to the LFS in 2021: https://statbel.fgov.be/en/changes-labour-force-survey-lfs-2021

Every year a lot of excel files are updated with the annual data (quarterly and yearly variables). These updates are associated with a press release about the main results and a press release about the flow statistics. No paper publications. Occasionally we put our LFS data in the spotlight, for instance on Labour Day (in 2021 about work organisation: https://statbel.fgov.be/en/themes/work-training/labour-market/focus-labour-market), Women's Day,...
9.3. Dissemination format - online database
Documentation, explanations, quality limitations, graphics etc.    
Web link to national methodological publication Conditions of access to data Accompanying information to data Further assistance available to users
 

Link to the national web page (national language(s)):

English: https://statbel.fgov.be/en/themes/work-training/labour-market/employment-and-unemployment

Dutch: https://statbel.fgov.be/nl/themas/werk-opleiding/arbeidsmarkt/werkgelegenheid-en-werkloosheid

French: https://statbel.fgov.be/fr/themes/emploi-formation/marche-du-travail/emploi-et-chomage


  Standard tables are available on our website http://statbel.fgov.be

Users can also make their own tables via our dynamic application be.STAT.
Other aggregated data can be made available on demand in excel-format without special conditions or costs. Also coded microdata can be delivered (with confidentiality contract). The procedure can be found on our website: https://statbel.fgov.be/en/microdata-research

The price of the data file is 500 euro.

 For results based on a small sample size we mention that they have to be interpreted very carefully. Confidence intervals are given on demand. Also extra information is given on demand.  We have introduced a FAQ page on our website: https://statbel.fgov.be/nl/themas/werk-opleiding/arbeidsmarkt/faq

There are some contact persons to help them if they have specific questions.
9.3.1. Data tables - consultations

[not requested for the LFS quality report]
9.4. Dissemination format - microdata access
Accessibility to LFS national microdata (Y/N) Who is entitled to the access (researchers, firms, institutions)? Conditions of access to data Accompanying information to data Further assistance available to users
 Y Researchers, firms, public services,... More information can be found on:
www.gegevensbeschermingsautoriteit.be		 The procedure can be found on our website:

https://statbel.fgov.be/en/microdata-research

The price of the data file is 500 euro.

 Codebooks + questionnaires Users can contact the statisticians for further assistance. For legal information users can send a

mail to statbel.datarequests@economie.fgov.be
9.5. Dissemination format - other

[not requested for the LFS quality report]
9.6. Documentation on methodology
References to methodological notes about the survey and its characteristics

https://statbel.fgov.be/sites/default/files/files/metadata/T8.STAT_DTST_21.CTAC_ORG_1.DIFF_LVL_1.NL.pdf

https://statbel.fgov.be/sites/default/files/files/metadata/SVY/T11.SVY_1.CTAC_ORG_1.DIFF_LVL_1.FR.pdf

 

A paper on the 2017 reform is published in 2018:

https://statbel.fgov.be/sites/default/files/Over_Statbel_FR/Analyse_eak_2017_nl_20181220.pdf

https://statbel.fgov.be/sites/default/files/Over_Statbel_FR/Analyse_eak_2017_fr_20181220.pdf
9.7. Quality management - documentation

[not requested for the LFS quality report]
9.7.1. Metadata completeness - rate

[not requested for the LFS quality report]
9.7.2. Metadata - consultations

[not requested for the LFS quality report]


10. Cost and Burden Top
Restricted from publication


11. Confidentiality Top
11.1. Confidentiality - policy

[not requested for the LFS quality report]
11.2. Confidentiality - data treatment
Please provide information on the policy for anonymizing microdata in your country
The dissemination of pseudonymised study data is strictly regulated. The procedure is described on our website: https://statbel.fgov.be/en/microdata-research. In order to get the permission of Statbel's Data Protection Officer team and finally as data controller, Statbel's director-general, the third party should follow a procedure and sufficiently motivate the proportionality and relevance of its request. The more confidential the information requested, the better the need for it should be motivated.


12. Comment Top

[not requested for the LFS quality report]


Related metadata Top


Annexes Top