Subsampling Callbacks for Difficult-to-Reach Respondents to Improve Survey Efficiency

Population surveys with multiple callbacks to maximize response rates are expensive, and often a substantial proportion of the data collection cost comes from the many callbacks required to obtain a small proportion of interviews with difficult-to-reach respondents. This research explores whether data collection costs are reduced by subsampling a random proportion of the originally sampled units from the mth callback attempt forward, with case weights utilized to adjust for the undersampling. The relationship to optimal allocation is also explored.