Do supplemental list frames for subpopulations increase subpopulation sampling efficiency? Evidence from the National Household Food Acquisition and Purchase Survey

Abstract

Multiple-frame sampling has been regarded as a device for increasing efficiency in identifying small subpopulations. However, there has been a lack of empirical evidence in supporting the efficiency of the multiple-frame approach and in guiding best practices. The current study focuses on a special scenario in which two frames were used to recruit sample members. Using paradata from the U.S. National Household Food Acquisition and Purchase Survey (FoodAPS), the current analysis focuses on recruiting households that received Supplementary Nutrition Assistance Program (SNAP) as a sub-goal of the survey sampling. SNAP households account for around one-fifth of the general U.S. population, compared to a survey goal of 30 percent of responding households. Our findings were consistent with theoretical expectations. Having and using additional SNAP list frames improved the efficiency of identifying SNAP households as opposed to screening a general address-based sample frame. This efficiency remained even as the SNAP list frames aged.