Two-Phase List-Assisted RDD Sampling

J. Michael Brick, David Judkins, Jill Montaquila, and David Morganstein

Abstract:
Random digit dial (RDD) telephone surveys are cost-effective means of sampling and interviewing households in the U.S. In many RDD telephone surveys, telephone numbers are sampled with equal probability using a list-assisted procedure. We consider an alternative two-phase approach to sampling telephone numbers. A first phase sample is selected and auxiliary data are used to stratify telephone numbers by whether they are listed or whether an address can be associated with the telephone number. The telephone numbers in the second phase strata are sampled at rates that balance data collection costs and the variances of the estimates. Our findings suggest that stratification by mailable status has greater benefits than stratification by listed status in RDD surveys. Methods for estimation and variance estimation for the two-phase design are also considered.

Keywords:
Optimal allocation; cost models; telephone surveys.