Alternative Imputation Methods for Wage Data

Sandra A. West, Shail J. Butani, and Michael Witt


This paper compares different methods of imputating establishment wage data. The methods include regression modeling and distribution modeling with maximum likelihood estimators for the parameters, as well as standard procedures such as ratio adjustment and hot deck. Multiple imputation procedures are also developed and examined. In all, ninety procedures, each using three sample designs, are examined. Wage and employment data used for this study are obtained from the Bureau of Labor Statistics census of establishments. Different nonresponse patterns are simulated. The effectiveness of the various imputation methods is measured both at the micro and macro level.