Abstract

When combining register and survey data, legal restrictions on record linkage can lead to missing values in register data variables for those persons who refused to give their permission. It is analyzed whether multiple imputation (MI) of the missing register data will avoid biased estimates due to systematically missing observations.