Weighting Cell Phone Surveys

Objective: To develop and evaluate a strategy for weighting survey data collected via landline and cell phones.

Introduction: Combining survey data from cell phone and landline interviews is difficult because people who have both landline and cell phones have an increased probability of selection into the combined sample, and current, reliable estimates of wireless penetration and substitution in many populations are not available.

Method: We developed a method for weighting combined survey samples based on the assumption that people with certain characteristics have knowable propensities to appear in landline and cell phone samples. Based on data from a national cell phone survey and from the National Health Interview Survey, we developed a propensity weighting model for combining cell phone and landline data.

Results: Results from the application of the model to data from a national, a state-level, and a regional survey suggest that propensity weighting effectively combines landline and cell phone data, effectively reproducing independent estimates of wireless penetration and substitution.

Conclusions: Propensity weighting provides an easy, valid means of combining landline and cell phone survey data for which independent estimates of wireless penetration are not available.