February 2000

IZA DP No. 117: IAB Employment Subsample 1975-1995 Opportunities for Analysis Provided by the Anonymised Subsample

Stefan Bender, Anette Haas, Christoph Klose

The IAB employment subsample is now available for researchers in a third, anonymised version. Following the so-called basic file and the regional file from the IAB employment subsample, which encompassed the years 1975 to 1990, the actualized version of the basic file covers now the years 1975 to 1995 and contains for the first time information on Eastern Germany for the period 1992 to 1995. Therefore, the IAB employment subsample is equipped with data of one percent of all employees registered by the social insurance system within the given period of 21 years. This data has been stored into a file by the Federal Employment Service (Bundesanstalt für Arbeit) to provide an insurance account for each employee recorded by the German social insurance system. Supplementary information on establishments and on unemployment periods in which a claimant received benefits was added to the sample. This version now contains exact daily flow information on the employment history of 559,540 persons as recorded by the social insurance system and on periods of drawing benefits as well. It allows to reproduce employment careers without typical problems of longitudinal surveys which do arise in social research (e.g. panel mortality, memory gaps). Nevertheless there are specific problems that appear as a result of the data generation process. In general, data from the employment statistics like the IAB employment subsample is subject to confidentiality under Social Code Book X. Passing this data to third parties would have made complicated approval procedures necessary. To avoid such difficulties the IAB decided on the anonymisation of the data with a procedure based on the factual anonymisation under the Federal Statistics Act. The project could only be realized with financial assistance from the scientific community. Along with the anonymising of person- and establishment-related cross-section information, the longitudinal information of the persons involved had to be anonymised factually, too. The procedure mainly consisted in the aggregation of the characteristics and in a shift of the complete employment history of each person on the time axis. Comparing the anonymised with the original IAB employment subsample it is shown that the anonymising procedures do not place any serious constraints on the analysis potential of the file. Though the distribution of establishment transitions along the time axis cannot not be reproduced with the anonymised IAB employment subsample, cross-section comparisons of stocks at an historical point in time and multivariate models aiming at daily accuracy (e.g. piecewise constant exponential models) do not lead to any deviations of the results between the two files. The anonymised IAB employment subsample, which is available to researchers through the Central Archive for Empirical Social Research at the University of Cologne, is, therefore, one of the most important German sources for employment market research.