Imputation Rules to Improve the Education Variable in the IAB Employment Subsample

The education variable in the IAB employment subsample has two shortcomings: missing values and inconsistencies in the reporting rule. We propose several deductive imputation procedures to improve the variable. They mainly use the multiple education information available in the data because employees' education is reported at least once a year. We compare the improved data from the different procedures and the original data in typical applications in labor economics: educational composition of employment and wage inequality. We find that correcting the education variable shows the educational attainment of the male labor force to be higher than measured with the original data and changes some estimates of wage inequality. Our analysis does not provide a definite rule on how to choose among the different imputation procedures discussed, but we recommend correcting the original education variable.