Exploring the Impact of Various Control Total Sources for Adjusting for the Cell-only Population in the California Health Interview Survey
Ismael Flores Cervantes
Westat
Matt Jans
UCLA Center for Health Policy Research
Trent D. Buskirk
Marketing Systems Group
Dual-frame (cell phone and landline) random digit dial (RDD) surveys greatly reduce coverage errors in a variety of survey estimates compared to landline-only RDD surveys. However, for state and local surveys, the choice of control total sources for post-stratification of the cell phone sample has been limited. Marketing Systems Group (MSG) produces quarterly county level estimates for proportion of cell phone-only households in a county based on phone number assignments and administrative databases. Since these estimates are produced for smaller geographical areas, their use can potentially improve estimates in dual-frame RDD surveys. We study the impact of using MSG county-level estimates as control totals for post-stratification weighting on estimates in California Health Interview Survey 2001-2012. Changes in the weighting process and qualities of the weights and estimates correlated with cell-only phone (CPO) use are addressed. Results show that although there are differences in the weights and estimates of CPO use, estimates produced using the new weights are not significantly different from estimates produced using the current weighting methodology at the state level.