Implementation of Controlled Selection in the National Compensation Survey Redesign

Lawrence R. Ernst, Christopher J. Guciardo, Yoel Izsak, Jonathan J. Lisic, and Chester H. Ponikowski


The paper "Update of the Redesign of the National Compensation Survey" (Izsak et al. 2005) included a proposal to allocate the number of sample establishments among cells using a controlled selection procedure, where cells are area PSUs x industry sampling strata x sampling panels. Since then the procedure has been implemented but with a number of modifications not discussed in Izsak et al. (2005). These modifications and possible future changes are discussed in this new paper. They include: weighting changes necessitated by the use of controlled selection, complications caused by rounding issues and how they were overcome, complexities caused by the need to allocate over five sampling panels, and use of a real-valued minimum allocation for each sampling cell in the controlled selection process in order to avoid very large sample weights and accompanying increases in variances.