James A. Buszuwski, Daniel J. Elmore, Lawrence R. Ernst, Michael K. Lettau, Lowell G. Mason, Steven P. Paben, and Chester H. Ponikowski (2003) "Imputation of Benefit Related Data for the National Compensation Survey."
This study examined imputation methods for benefits data collected as part of the integrated National Compensation Survey (NCS) program, which is conducted by the Bureau of Labor Statistics (BLS). Among the components of NCS are the Employment Cost Index (ECI), which measures changes in wages and benefits via a Laspyres index; the Employer Cost for Employee Compensation (ECEC), which measures mean employer costs for wages and benefits; and the Employee Benefits Survey (EBS), which studies the incidence and plan provisions of employer-provided benefits. The integration of these surveys will allow NCS to link employer benefit costs to benefit incidence and plan provisions and calculate new measures, such as cost per participant. As part of this paper, we revisit the imputation methods used prior to the integration; evaluate their use as part of NCS; and introduce methods for handling item nonresponse for the new measures. We suggest regression methods for handling all missing cost related data and the nearest neighbor within-cell hot deck method for handling all missing incidence and provisions data.
Last Modified Date: July 19, 2008