Series: Microcensus (MZ)

Due to the federal organization of the German statistical system, the federal statistical office cooperates with the statistical offices of the Länder. The planning and preparation of the Microcensus is carried out by the Federal Statistical Office. The statistical offices of the Länder are responsible for conducting surveys and preparing the data.

Sampling method

Single-stage stratified cluster sampling (since 1972)

Sampling fraction

Core program: 1 % of the residential population

2005-2011: Ad hoc modules 0.1 %; EU labour force survey structural variables 0.1 %

2012-2019: Ad hoc modules 0.1 %; EU labour force survey structural variables 1 %

2020: Labour Force Survey (LFS) maximum 45 % of total sample; Statistics on Income and Living Conditions (SILC) maximum 12 % of total sample (collected separately up to and including 2019).
Beginning in 2021: Labour Force Survey (LFS) maximum 45 % of total sample; Statistics on Income and Living Conditions (SILC) maximum 12 % of total sample (collected separately up to and including 2019); Information and Communication Technologies (ICT) in households maximum 3,5 % of total sample (collected separately up to and including 2020).

Sampling frame

Up to 2015 the master sample for the old federal states was based on results of the 1987 Census of Population and for the new federal states and East Berlin on results from the "Bevölkerungsregister Statistik 1990" (statistics derived from the central population register) of the German Democratic Republic (GDR). From 2016 onwards the 2011 Census serves as selection basis for the Microcensus. The master sample is updated annually by taking new construction activities into account. Since 1990, updating has been carried out by drawing a new-construction sample from building activity statistics. In the 1972–1989 rounds of the Microcensus, the supplementary sample was realised either on the basis of interviewers' reports of new construction activity in the sampled municipalities or on the basis of zoning maps.

Primary sampling units

The sampling districts (primary sampling units) are artificially delimited areas (clusters) comprising, as a rule, neighbouring buildings; in the case of larger buildings, only one building, or a subdivision thereof, is included. Up to the 1989 Microcensus, a sampling district comprised, on average, 23 dwellings. To enhance regionalisability, the average size of the sampling districts was reduced in 1990. Since then, the average sampling district comprises 9 dwellings. The target size for sampling districts in the collective-household building stratum is 15 persons.

Partial rotation

1974-1977: Rotation halves remained in the sample for two years; each year half of the sampling districts were replaced.

1977-2019: Each sampling district remains in the sample for four years. Each year, a quarter of the sampling districts are replaced by newly introduced sampling districts. 

Since 2020: Sampling districts can be surveyed up to four times. Each year, approximately 30 percent of the sampling districts drop out, while a correspondingly large proportion rotate into the survey. Private households in sampling districts that are part of the LFS subsample are surveyed repeatedly during the year (surveyed repeatedly for two consecutive quarters, then not surveyed for two quarters, then surveyed again for two consecutive quarters). These households thus rotate out of the current sample more quickly. Institutional households in LFS sampling districts are interviewed only in the first and third survey waves and in the fourth if it falls in a first calendar quarter.

Households and persons moving out of the sampling district are not followed but rather replaced by households and persons moving into the sampling district. The Microcensus is therefore a repeat survey with partial overlapping of sampling units. The rotation quarter to which a respondent belongs is not identifiable in the basic and regional Scientific Use Files (SUFs). Since 1996, it has been legally possible to aggregate the cross-sectional data to form a panel data set. Two panel data sets are available as SUFs for panel analyses: 1996-1999 and 2001-2004. A new procedure for the subsampling and time-consistent identifiers allows researchers to independently create panel files from the Scientific Use File 2012 onwards (see also: Microcensus Scientific Use Files from 2012 onward as a Rotational Panel). As a qualification, it needs to be mentioned that a new sample was drawn in 2016, so that panel files can be created only from 2012 up to 2015. Furthermore, panel files from 2016 onwards are limited to the period up to 2019, as the Microcencus and other household statistics have been converted to a fundamentally new system from 2020 onward.

Guidelines for the Factual Anonymization of Particular Statistical Microcensus Information (From the appendix to the standard contract of the Federal Statistical Office regarding the dissemination of Microcensus data)

1. In accordance with the Microcensus Act, direct identifiers, such as personal identification numbers, names, addresses, telephone numbers, or any other official numbers that permit identification of persons from the data may not be distributed. Therefore, they may not be contained in the data file.
2. The dataset must be transmitted in an unsystematic order.
3. Only a randomly generated 70% subsample may be distributed.
4. The only direct regional data to be given are those of the respective German state.
5. In the "community size" category, no individual community with a population of less than 500,000 may be identifiable. If there are multiple communities in the same category, they must together comprise at least 400,000 inhabitants in every German state.
6. The citizenship or nationality of groups in the Federal Republic of Germany with less than approximately 50,000 persons are not to be identifiable.
7. In its univariate distribution within the statistical population, each category trait must include an aggregate of at least 5,000 cases. If this precondition is not met, then a statistically appropriate simplification or aggregation is needed in order to reach the required number of cases.

To item 3: The Microcensus Scientific Use File (SUF) is a de-facto anonymised 70 percent subsample of the households in the Microcensus. Until 2011, households or dwellings were primary sampling units of the Microcensus, with all persons of a selected household or dwelling being included in the sub-sample. Dwellings were drawn in the survey years in which the 4-year supplementary programme on housing (1998, ..., 2010) was carried out. The basis for the systematic random selection is the original data of the Federal Statistical Office. From 2012, the primary sampling units for the sub-sample are the sampling districts within a rotational quarter.

Extrapolation in the Microcensus follows a two-stage procedure. The first stage – correction, or compensation, for non-response – entails extrapolating from the net sample of successfully surveyed households to the gross sample of all households on the basis of the design weight. In the second stage – adjustment – the weighted sample is adjusted to conform to benchmark figures from the intercensal population estimates (laufende Bevölkerungsfortschreibung). Depending on the methods used, the two-stage procedure yields different extrapolation factors for the individual survey years.

The integrated EU-Labour Force survey (LFS) the Survey on Income and Living Conditions (SILC), which has been integrated since 2020, and the Survey on ICT Usage in Households and by Individuals, which has been integrated since 2021, allow for comparisons with other EU member states. The Microcensus data allows for comparisons at the level of federal states (Bundesländer) as well as more fine grained regional units.