An Empirical Comparison of Two Telephone Sample Designs

The clustered telephone sample design described by Waksberg is compared with a design randomly generating four digit numbers within working prefixes. The clustered sample is found to increase the proportion of working household numbers selected from about 22% to over 55%, but sampling errors and design effects of the two sample designs show some loss of precision in the clustered design. A cost-variance model is constructed which provides estimates of desirable cluster sizes given varying amounts of intracluster homogeneity.