Adjusting the German Labor Force Survey for Under-Reporting by Means of Calibrated Imputation

• Activity Number: 35
• Type: Contributed
• Date/Time: Sunday, July 29, 2007 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Hans Kiesl*+ and Susanne Rässler
• Companies: Institute for Employment Research and Institute for Employment Research
• Keywords: calibrated imputation ; measurement error ; labour force survey ; multiple imputation

Abstract

The German LFS, which is conducted as a CAPI household survey (including proxy interviews), suffers from serious under-reporting of marginal employment. As a result, the number of unemployed is overestimated. To adjust the LFS for this kind of measurement error, we use auxiliary data from administrative files to develop a calibrated imputation algorithm. First, we estimate provisional propensity scores for marginal employment by means of a binary choice model. Then, we change these propensities so that they agree with auxiliary totals from administrative data and at the same time minimize some distance function between the provisional and final propensities. In the last step, we impute a (calibrated) binary variable indicating marginal employment. To get valid variance estimates, multiple imputation is performed. The German NSI intends to use our procedure from May 2007 onward.