The EU Labour Force Survey (EU-LFS) is the largest European household sample survey. Its main statistical objective is to classify the population of working age (15 years and over) into three mutually exclusive and exhaustive groups: employed persons, unemployed persons, which together represent the ‘labour force’, and the people outside the labour force.

Population concept 

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, including people living in institutions

 Family home

 Term address (if private)

 Family home

 Family home

 depends on 1. whether or not the childred are registered at the adress and 2. are perceived as part of the household by the adult members.

The main users cover the entire spectrum of users of data from official statistics, including governments, international organisations, employers’ associations and trade unions, researchers, media and the general public, all at national as well as regional level. Institutional users include national and regional governments (e.g. the ministries of labour and social affairs), the German Federal Bank, the Federal Employment Agency, the European Commission (e.g. Eurostat and DG employment, social affairs and inclusion), the International Labour Organisation (ILO), the Organisation for Economic Co-operation and Development (OECD), the European Central Bank (ECB) and other international organisations. Also foreign governments frequently use LFS data for international comparisons. LFS data are widely used by employers’ associations, trade unions, trade associations, and individual enterprises both at national or regional level, for their decision making. Market, social and economic research companies use LFS results both as a basis for analysis and as input for weighting and calibration of commercial surveys. Universities and research institutions make frequent use of LFS data both in form of tables provided by the Federal Statistical Office upon request and via micro data access for research purposes made available via the research data centres of official statistics. Results from LFS data are finally frequently disseminated by general and specialised media, both at national and at regional level and requested by the general public. In 2016, more than 2,000 user requests were received from these user groups. Since 2016 the number of requests can not be withdrawn from the software, as the information service is centralized within the statistical office. Furthermore, the number of request is reducing as more information can be found online.

The data requirements of the users are taken into consideration via different channels and are mirrored in national legislation - including the LFS. The ministries of the Federation and the Federal States can directly influence the list of variables included in the Microcensus through national legislation. The sub-committee “Employment/Labour Market” of the German Statistical Council brings together representatives from major user groups including government ministries, employers’ associations and trade unions, trade associations and chambers, environmental and conservation organisations, universities, the head organisations of the municipalities and the council for social and economic data. The sub-committee “Employment/Labour market” meets biennially and provides users a forum in which new developments are being presented and where new or changed data requirements can be discussed.

(Y/N)

Give a description of difference and provide an assessment of the impact of the divergence on the statistics

 Y

 In the German LFS only people living in dwellings are part of the frame. Homeless people and other people without registered residence (e.g. people living in huts, caravans) are out of the frame.

 N

 N

 N

Total employment

 Employed persons living in collective households (not in the LFS); Treatment of persons with a job on long-term absence (only included in LFS if duration of absence <3 months or continued receipt of at least 50% of employment-related income);  Employees below the age threshold of the LFS. Domestic and national concept

 Measurement differences compared to sources used by National Accounts (LFS underestimation of marginal part-time employment); Adjustments in National Accounts (informal employment)

Labour Force Survey result (age 15+) shows less employed persons than national accounts estimates. national concept -3.4 million, domestic concept -3.6 million

 Differences between employment figures of Labour Force Survey and national accounts estimates - German Federal Statistical Office (at: link to publication)

Total employment by NACE

 The National Accounts use the enterprise concept, the LFS the local unit/establishment concept; differences regarding employment, see above. Domestic concept. 

 Methodological differences (LFS captures industry as indicated by respondents; National Accounts rely on register information).

 Labour Force Survey results underestimate the Agricultural -0.06 million and Service Sector -3.9 million and overestimate industry (except construction) +0.3 million

 Körner, Thomas and Katharina Puch: Coherance of German Labour Market Statistics. In: Statistik und Wissenschaft, Band 19, 2011. Link to publication

Number of hours worked

 No conceptual significant differences regarding the concept of hours actually worked; differences regarding employment, see above. Domestic concept. 

 The calculation of hours worked within the framework of the National Accounts is based on a differentiated component wise accounting concept, where calendar effects, collectively agreed standards, business cycle influences as well as personal and other components are considered. The accounting model uses a total of 20 different statistics, including the LFS (which nevertheless plays only a limited role); LFS underestimation of marginal part-time employment; LFS underestimation of absences during the reference week, e.g. due to holidays

In contrast to many other countries, the working time estimation of National Accounts makes only limited use of LFS data in the German case (the most important use being the estimation of the working time of the self-employed). The estimation of working time

for National Accounts is based on a differentiated component wise accounting approach that combines 20 different statistics (see Susanne Wanger, Tobias Hartl, Markus Hummel, Yasemin Yilmaz: Überarbeitung der IAB-Arbeitszeitrechnung im Rahmen der Generalrevision 2024 der Volkswirtschaftlichen Gesamtrechnungen. In: IAB-Forschungsbericht 20,2024. Available at Link to publication ).

 Körner, Thomas and Loup Wolff: Do the Germans really work six weeks more than the French ? – Measuring working time with the Labour Force Survey in France and Germany. In: Journal of Official Statistics 32. Link to publication

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 doesn’t 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. employ-

ment registers and/or enterprise

surveys)

 N

 N

 Y

 N

 N

 Y

Total employment

 NA

 NA

 NA

 NA

Total employment by NACE

 NA

 NA

 NA

 NA

Number of hours worked

 NA

 NA

 NA

 NA

 Criteria for ILO-unemployment and registered unemployment are different. Registered unemployed can have a job with less than 15 working hours; then in compliance with ILO they have to be classified as employed. Criteria for registered unemployment are: Working less then 15 hours per week, registered at the employment agency, available for the employment agency. Registration at the employment agency is not an ILO-criterion.

 Registered unemployed can only be persons who registered with the public employment agency. ILO-unemployment is measured by a survey; it is independent of being registered or not.
Unemployed persons, 65 years and older, who are looking for a job, cannot be registered. The actively search for a job, which is a condition for both statistics to be counted or registered, is construed in different manners

 Körner, Thomas and Katharina Puch: "Der Mikrozensus im Vergleich mit anderen Arbeitsmarktstatistiken. Ergebnisunterschiede und ihre Hintergründe seit 2011." In: Wirtschaft und Staitsik 4/2015.
Available at: Link to publication

Overall effect

Men under 25 years

Men 25 years and over

Women under 25 years

Women 25 years and over

Regional distribution (NUTS-3)

'- 1.3 million (less unemployed than registered)

15-64 years

 + 0.02 million (more unemployed than registered)

 '- 0.68 million (less unemployed than registered)
15-64 years

 + 0.01 million (more unemployed than registered)

 ' -0.64 million (less unemployed than registered)
15-64 years

  na

 Final sampling unit (FSU)

Date of sample selection

 The basic concept of sampling methodology is one-stage cluster sampling (area sampling). Explanations on the rotation design applied are given below.

 The census data from 2011 is the survey base for the sample.

 The annual update of the sampling frame is realized by using information on building licences. These information was available in all Federal States in autumn 2017.

 Sampling districts consisting of 9 dwellings on average (area sampling). All buildings are assigned to one of three size classes, depending on the number of dwellings they comprise. Buildings with less than 5 dwellings belong to the first size class. In this size class, each sampling district comprises 12 dwellings on average (usually in neighbouring houses in rural areas). The second size class comprises medium-sized buildings with five to 10 dwellings. Each of these buildings constitutes a sampling district. The buildings in the third size class comprise 11 dwellings or more. In this size class sampling districts are subdivisions of the building, the target size being 6 dwellings. An additional stratum covers the population living in collective households. Collective households are divided into sampling units with a target size of 15 persons. New buildings reported are allocated to the size classes specified above. Compared with the selection based on the 1987 population census, the following modifications have been made: The sample districts formed by buildings with 1 to 4 dwellings have a target number of 6 dwellings (instead of 12). The minimum number of dwellings per building is 9 in the third size class. This means that the sample districts of all building classes have roughly the same size.

Stratification (variable used)

Number of strata (if strata change quarterly, refer to Q4).

Rotation scheme (2-2-2, 5, 6, etc.)

 The sampling method applied, i.e. sorting, defining zones, and sampling within the zones, guaranteed an effect similar to stratification. The regional strata (see "stratification") were grouped to 131 "adjustment strata", at least to the extent that an average of 500,000 inhabitants was obtained. Bound expansion is performed at that regional level. Before sampling, the sample districts were sorted in terms of region, i.e. within each stratum they were sorted by regional stratum subgroup, administrative district, community size class, community, and sample district number. "Zones" were formed by 100 consecutive sample districts each. The sample districts of each zone were formed at random by permutation of numbers 0 to 99 by means of a random number generator. Sample districts with the same number, i.e. the same "sampling number", were grouped to form a 1%-sample. Thus the population was divided into 100 1% samples. The random number generator was also used to form at random four successive zones each by permutation of numbers 1 to 4. This permitted to divide every 1% sample into 4 rotation quarters of 0.25%. The 20 1% stock samples were determined at random by sampling from an urn an interval comprising 20 sampling numbers between 0 and 99. Subsequently, the first 1% sample to be used was determined also by sampling from an urn. The subsample for the yearly variables, too, are obtained systematically with a random start.

 All households in the sampling districts and all persons in the housholds are to be surveyed.

 Stratification is done for the overall microcensus sample and applies also for each subsample (including the LFS):
The sampling districts are stratified by region and size of buildings. The stratification by size of buildings is based on the size classes used to work out the sampling units. There are 243 regions which comprise 200.000 inhabitants on average. The sampling rate is the same in each stratum. The list of sampling districts is sorted within each stratum by sub-region, Kreis (administrative district), the size class of the commune, commune and number of sample district. Within each stratum, an effect similar to stratification is obtained by systematic sampling in a list classified by geographical entity. The list of sampling districts in each stratum is devided into groups of 100 consecutive sampling districts. In each of these groups a number of sampling districts in accordance with the regional sampling ratio – which average to a 1% sample for the microcensus (and roughly 0,4% for the LFS) on the national level - 1% is drawn at random in each of these groups. In order to account for the differing design effects of the individual NUTS2 regions the regional sampling ratio do vary quite a bit. The 1%-sample is allocated to the months and quarters of the year at random. The sample is refreshed each year by sampling districts of new buildings. The building size is utilized in form sample districts, but due to its low numbers they form just one subject-related stratum (instead of 4) per regional stratum.

 131

 The rotation system is composed of four waves and the rotation scheme is 2-2-2. Each sampling district remains in the sample either for 2 or 3 years (2 years if the first wave is during Q1-Q3). 4/9 (44,4%) of the LFS-sample districts is replaced each year. Thus, the degree of overlapping between two consecutive yearly samples is 5/9 (55,6%).

 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. 2019/2240) (Y/N)

If not please list deviations

List of yearly variables for which the wave approach is used (Ref.: Commission Reg. 2019/2240, Annex I)

 1st and 3rd wave and 4th wave if the reference week is in Q1.  Households in the 4th wave and Q1 are surveyed  because the German microcensus law states that these structural variables are to be gathered from each LFS household each year.

Y 

Consistency of ILO labour status in combination with age-brackets and gender between the yearly and quarterly sub-samples has to be met by the extrapolation of the LFS data.

 NA

 Y

Before data can be compiled, in a first step all households living in a sampling district must be identified. At the beginning of the week following the reference week, sampled housholds are informed via a letter that they form part of the microcensus sample. They are given the choice to participate by way of CAWI (login information included in the letter) or a telephone interview conducted by an interviewer or a clerk of the regional statistical institute. If neither option fits the respondent’s situation than a PAPI is offered. In some cases, an interviewer can visit the respondent at his welling and conduct the survey via CAPI.

If the sample household has not responded after approx. 2-3 weeks, a letter reminding him of his obligation to participate in the survey will be sent. As long as the respondent does not react and participate, he/she will continue to receive reminders – eventually introducing and escalating fines – until he/she complies.

When conducting the survey via an electronical mode (CAWI,CATI, CAPI) each entry will trigger plausibility checks and respondents are only able to advance in the questionnaire if all erroneous entries are corrected. If the responded has conducted the survey via paper questionnaire then the same plausibility checks will be applied after the data submitted is copied into the electronic entry mask. In case of missing data the statistical institute will reach out to the respondent.

When the regional statistical institute clears the compiled data of a respondent household in order to granting access to the national statistical institute once again several checks for plausibility and completeness of the data are applied.

 Y

 Self-administered software solution

 N

A two-stage adjustment procedure is used:

(a) Due to Covid-19 and technical issue it was not possible to access sufficient information on households who failed to respond.  As a proxy the structure of 2019 LFS-sample was adjusted by regional sampling ratios of 2020. The net-sample was calibrated to this proxy based on education (low, medium, high), nationality (German, non-German), age (65 and under, over 65).  and household size (1person or more). Non-Response weights were calculated as the inverse of the received calibration weight. 

(b) The final sampling weights are calculated with GREG-estimator using auxiliary variables compromised of combinations in different dimensions of nationality, sex, age group, region NUTS-2, employment status

 Y

 NA

 Y

0-14,15-24,25-34, ..., 65-74,75+

 NUTS2

 Education, nationality, employment status

A two-stage adjustment procedure is used:

(a) If possible, limited data are collected for households which fail to respond. The weighting of responding households to take account for non response includes calibration towards number of persons of the household, nationality (German/foreign) and for households comprising only one person: age (under or over 60 years) and sex. The weighting is part of a composite estimation with the second part being log.regression of response rates of previous years. Due to Covid-19 and technical issue it was not always possible to access sufficient information on households who failed to respond.  As a proxy the structure of 2019 LFS-sample was adjusted by regional sampling ratios of 2020. The net-sample was calibrated to this proxy as well as the response rates for 2021 based on education (low, medium, high), nationality (German, non-German), age (65 and under, over 65)  and household size (1person or more). Final Non-Response weights were calculated as the inverse of the received composite estimation weight.

(b) The final sampling weights are calculated with GREG-estimator using auxiliary variables compromised of combinations in different dimensions of nationality, sex, age group, region NUTS-2, employment status

 Y

 Y

 NUTS2

 Education, nationality, employment status, number of private households, size of household

 N