Redesigned microcensus as of 2020 and methodological effects
The microcensus has been redesigned. The European labour force survey (LFS) has long been integrated into the microcensus. The European Union statistics on income and living conditions (EU-SILC) has also been part of the microcensus since survey year 2020 after having been compiled by a separate survey. Since survey year 2021, the survey of the use of information and communication technologies (ICT) in private households has also been included in the microcensus. These new components offer many new analysis options.
The redesign also involved changes to other features of the microcensus. The list of questions and the concept of the sample survey were modified, and with the introduction of an online questionnaire, the form of data collection were also changed. The results from survey year 2020 onward are therefore comparable with those of previous years only to a limited extent.
Detailed information on microcensus modifications from 2020 is contained in the paper entitled "Die Neuregelung des Mikrozensus ab 2020" (Rearranged microcensus as from 2020), which was published in the 6/2019 issue of the Wirtschaft und Statistik (WISTA) scientific journal.
Impact of the redesign and the Covid-19 crisis on the microcensus 2020 in general
A completely new IT system was set up for the redesigned microcensus; its introduction was accompanied by technical problems that affected the survey conduct.
The situation was aggravated by the pandemic in 2020, which made it nearly impossible to interview respondents face-to-face on the spot as had been common practice in the years before. Social distancing and stay at home orders had an effect not only on the face-to-face interviews in the households but also made preliminary on-site visits more difficult, which are necessary for sample specification. In addition to that, the official statistics agencies mostly refrained from sending out reminders to account for the other pandemic burdens the population had to bear. This means that the obligation to provide information was not as usually enforced. All these factors led to a lower response rate than is common for the microcensus.
Due to these problems, the average non-response rate for final results of the 2020 microcensus is approximately 35 % (first results approx. 38 %) at federal level, which is markedly higher than in previous years. This non-response was reacted to on the basis of a mathematical-statistical model.
More precisely, this means that non-response is distributed very unevenly across subject areas, regions and periods of time. Any distortions that may occur are therefore countered by calculating response probabilities before extrapolation (calibration to the key figures of population statistics). Information on the gross sample is needed for that purpose. As it was not possible either to cover the gross sample completely in 2020 (interviewers could make preliminary on-site visits to a limited extent only), a synthetic (artificial) adjustment frame was created to compensate for that. It was assumed that the 2019 sample distribution corresponds to the gross sample distribution of 2020 (the variables remain stable over that period). The usual procedure (calibration of the net sample using the gross sample) could thus be followed to calculate response probabilities for the 2020 microcensus at the level of the Länder. This improves the quality of the results.
The trade-off between the reliability of results and our publication practice has led to a conservative release policy in this case. Therefore, the usual technical and regional level of detail of the microcensus results cannot be reached due to the specific situation in 2020 described above.
More information about the redesigned microzensus.
Impact on Labour Market Results
As the pandemic has changed the situation on the German labour market considerably in many sectors, there is additional uncertainty as to how labour market results should be evaluated when they are released. Statements as to whether results are attributable to actual developments or to the methodological and technical limitations described above can thus be made only with certain reservations. These uncertainties increase with the level of detail (for unemployment rates with detailed regional or demographic breakdown, for example).
Analyses of the quality of atypical employment data have revealed implausibilities which made imputations necessary. It is likely that the wording of the question on temporary employment was misunderstood by many respondents who filled in paper or online questionnaires on their own without interviewers providing explanations. A number of respondents answered yes to that question although they were not in temporary employment. For respondents who had completed an online or paper questionnaire in 2020 and 2021, the answer to the question about temporary employment was therefore imputed by means of a correction procedure (hot deck procedure). Here the probability that respondents are in temporary employment is derived from the response behaviour in the other interviewer supported modes of data collection (face-to-face, telephone interview) for the imputed values. Data on temporary employment were also partially imputed in 2022, as the adjustment to an improved questionnaire has not yet been completed for all survey modes.
Apart from the persons in atypical employment there are effects on several other the labour market results. For example, the self-employed show a decrease of 10% in comparison to 2019. The reasons for this decrease are not easy to identify. It seems that a number of changes and effects, in addition to the pandemic, influences the measurement and the results.