The household component of the Community Tracking Survey (CTS) recently finished its third wave of data collection. The household survey employs a clustered survey to provide local estimates and an unclustered RDD survey to increase the precision of national estimates. Two features of the clustered sample are large (over 1000 households) samples in 12 of 60 sites (PSUs) to allow for precise local estimates and an inperson supplement in those same 12 sites to increase frame coverage.
This paper reviews the sample selection and weighting strategies over the first three waves of the CTS: how they accommodated the needs for precise estimate of change while maintaining precise and unbiased cross-sectional estimates, and controlling survey costs.
The design included selection at the second and third rounds of telephone numbers and households selected in the previous round. In the third round, optimal allocation based on cost led to oversampling households interviewed at the time of the second round.
|