To improve the accuracy of survey data regarding the responders’ ILO employment status, in 2005 and 2006 some questions about participation in the labour market, as well as fieldwork methods, were modified. Thus, any differences in distribution may correspond to changes on the labour market or, alternatively, may be due to the modification in the method (new questions). A comparison between the years 2006 and 2005 shows that certain questions are worded differently and/or include different possible answer categories. This applies to the following variables:

Drawing subsamples

• The procedure for drawing subsamples as used in 2002 and 1998 was modified for the 2006 survey to take into account the fact that some households in the 2005 survey were only questioned in 2006 and their answers included in the Microcensus sample for the first time. Using the general method of a 70% housing unit subsampling, these persons and households would be excluded as part of the so-called "Jahresüberhang" (i.e. households or persons who could not be interviewed in 2005, mainly because they could not be reached during the survey period and who were therefore not be interviewed until 2006). However, these units are needed for projection of features of the basic programme of the 2006 Microcensus. Thus, in order to give proper weight to aspects such as housing unit, household and personal evaluations, firstly a housing unit subsample was taken from households with no "Jahresüberhang", and secondly a supplementary household subsample was taken from households with a "Jahresüberhang" (see below).
• The Scientific Use File is a factually anonymized 70% subsample. In contrast to the other surveys, in the additional programme’s survey years with questions about the responders’ living situation, no household subsample is taken; rather, a 70% subsample of the housing units is carried out. Thus, in order to give proper weight to aspects such as housing unit, household and personal evaluations, firstly a housing unit subsample was taken from households with no "Jahresüberhang", and secondly a supplementary household subsample was taken from households with a "Jahresüberhang". This procedure is largely based on the sampling process of the Microcensus.
• The housing unit subsample (level 1: households and persons, and persons in shared housing units without a "Jahresüberhang") is a random sample systematically taken from the originally obtained data. This data is sorted according to reporting annual quarter, federal state, administrative district, community size, number of persons in the housing unit, classified size of the building, number of the district selected, building number and housing-unit number. (The size of the building was included as a feature for the first time in 2006.) This order ensures that the subsample will contain only slight random deviation from the original data as regards the abovementioned features.
• In the case of shared housing units, each person is assigned his or her own consecutive housing unit number. The last (single) digit in the housing unit number is used for sample drawing. All housing units whose single digit does not match one of three randomly selected one-digit numbers are used for the sample (70% of the housing units).
• The supplementary household subsample (level 2: households and persons, and persons in shared housing units with a "Jahresüberhang") is also a random sample systematically taken from the originally obtained data. The data is sorted according to reporting annual quarter, federal state, administrative district, community size, number of the district selected, and household number, in a method largely analogous to the other scientific use files without the additional programme addressing the responders’ living situation. The procedure for shared housing units corresponds to the procedure described above, i.e. each person is assigned his or her own consecutive household number. In the same way, the last (single) digit of the household number is used for the drawing of the subsample. Using the last digit procedure, households whose single digit does not match one of three randomly selected one-digit numbers are used for the sample (70% of households).

"Jahresüberhänge"

• The 2006 results contain so-called "Jahresüberhänge", i.e. data obtained from persons who could not be questioned in 2005 (due to lack of availability or other reasons). Regarding questions about the responders’ living situation which were only posed in 2006, this group (EF5b=1) must be excluded from analyses conducted for the 2006 survey period. However, the "Jahresüberhänge" must be used when evaluating the features of the basic programme, so that the published results of the statistical offices can be replicated.

Survey questions

• No final data cleansing was performed for certain derived variables (typologies) from the 2006 Microcensus. Thus, they may contain inconsistencies. These variables are marked with a star (*) in the index (see table, in German).
• For the identifiers and/or ordering numbers for the household (EF4), the building (EF8) and the housing unit (EF9), households with a "Jahresüberhang" from the 2005 survey (EF5b=1) were coded using their own ordering numbers, independently of the 2006 survey. As a supplement to this, the SUF also includes ordering numbers in which households providing data for the 2005 survey (ef5b=1) and the 2006 survey (ef5b=0) have identical codes (see EF4b, EF8b, EF9b).

• Attention: In the 2005 and 2006 Microcensuses, the federal states (EF1) are coded by alphabetical order; in the Microcensuses conducted up to 2004 and after 2007, the codes were assigned in accordance with the official community identifying number.

• Nationalities are recorded with a greater degree of specificity in 2006 than in 2005 (see EF369ff.).

• Recoding of Double Identifiers in the SUF 2006] [.do] [.spss] [.sas]

• Microcensus Scientific Use File 2006 Documentation and Data Preparation [in German] [.pdf]

• Persons (in private households and collective dwellings)
• Households
• Dwellings

• Persons
• Living arrangements
• Families
• Households
• Dwellings

01-01-2006

31-12-2006

Yes

• 76.9 % CAPI
• 2.5 % CATI
• 20.6 % Self-administered

25.53 %

Dwelling, Household, Persons

The Microcensus is a single-stage, stratified cluster sample with a uniform sampling fraction for all strata. As a rule, the sampling fraction is 1 %; it applies to all sampling units (dwellings, households, persons).

Because the selection probability of the Scientific Use File (SUF) basically remains constant at 70%, design weights can be created on the basis of the inverse of the sampling fraction of the Microcensus (1 %) and the Scientific Use File (70 %): w = 1 / (0.01 * 0.7).

In the Microcensus, as in all surveys, the sampling plan cannot be realised fully. Hence, undercoverage occurs in the form of non-response on the part of the households to be surveyed. Because households and their members are obliged by law to provide information for the Microcensus, non-response is due mainly to the fact that households could not be reached during the survey. Cases of non-response are corrected using the information that is available about the non-responding households.

Since 2005, the compensation factors are no longer calculated by dividing target figures by actual figures but rather by means of generalised regression estimation (calibration technique). At regional level, the following characteristics, or combinations of characteristics, are taken into account for private households: rotation quarter and new-construction stratum (federal state), 19 compensation classes (see above; regional adjustment stratum), and total private households (regional subgroup). For collective dwellings: number of persons in collective dwellings (regional subgroup).

In contrast to compensation, adjustment is carried out at person level.

Since 2005, adjustment has no longer been carried out at the level of disjointed adjustment strata but rather on a quarterly basis at different regional levels; data from the Central Register of Foreign Nationals (Ausländerzentralregister) are also used. Moreover, the same extrapolation factors are available for all members of a household. The extrapolation terms and regional units are: age groups x sex; nationality groups x sex; population group (temporary career volunteers and regular soldiers including members of the Federal Police and the Bereitschaftspolizei (riot police), persons doing military service, civilian population) (federal state); nationality groups x sex (administrative district); total population (regional adjustment stratum).

For the final weight a two-stage procedure is employed: In a first step the net sample of the successfully interviewed households is adjusted to the gross sample of all households to be interviewed (compensation). In a second step this gross sample is benchmarked to the update of the current population figures (adjustment). The final weights result from a combination of this two-stage procedure of compensation and adjustment. They are derived from the original data, are already multiplied with the inverse of the selection probability and extrapolated to 1000 of the population.

From 2005 onwards, in addition to the standard weighting factors for annual averages there also exist weighting factors for the quarterly samples, for the ad hoc modules and structural variables of the EU-LFS. Moreover, there are special weighting factors for supplementary programmes of the Microcensus (e.g., on the housing situation).

See [.pdf] (in German)