Optimal Allocation For Stratified Telephone Survey Designs

Robert J. Casady and James M. Lepkowski


Two stage random digit dialing (RDD) telephone sample designs such as that described by Waksberg are highly effective at identifying banks of numbers that have high concentrations of residential telephone numbers. Thus, they are able to increase dialing efficiency in telephone surveys by reducing the number of nonworking numbers that are dialed. Unfortunately, two stage RDD designs have undesirable features such as replacement of nonworking numbers late in the interviewing period. Hybrid telephone sample designs are proposed which provide samples of telephone commercial lists of listed telephone numbers to identify banks of numbers that have high concentrations. Allocation of sample size among listed numbers, unlisted numbers in banks with listed numbers, and numbers in banks without listed numbers are explored. Analysis of a commercial frame of listed numbers are used to specify design parameters in the allocation formulation.