Abstract

Respondents and Nonrespondents in three different population surveys will be compared using complete registry individual data on all designated respondents. The comparison will include not only age and sex, but nationality, start and end of each officially known job spell, the ISCO-code of the occupation, the daily wage, schooling and training, the industry and regional code of the working place. For the unemployed receiving social benefits, additional comparisons will include marital status, presence of children, type and amount of the benefit, start and end of each known unemployment spell. Detailed univariate comparisons of respondents and nonrespondents will be done and multivariate  models for the explanation of nonrespondents will be fitted. Furthermore, imputed and differently weighted data sets of the incomplete data sets will be compared to the data set augmented by data on nonrespondents. So we will be able to compare the performance of many different suggested nonresponse compensating models, including panel models, with actual data on nonrespondents.