Summary of Empirical Results of Two Sample Design Options for Sampling for the Count
C. T. Isaki, J. H. Tsay, Y. Thibaudeau
RR 93/05
ABSTRACT
As part of research on investigating alternative scenarios for the design of the next Decennial
Census, we looked into the possibility of sampling from the m ail universe in order to provide
voting rights data. Unlike previous censuses in which every address in the mail universe is
mailed a form, we developed several survey designs for estimating block, ARA and district office
totals. Using 1990 census data, empirical results on bias and mean square error were produced
for four district offices. We considered both a cluster design of addresses using blocks, and a
design that selects addresses in every block. The results were not encouraging.