A Design To Oversample Low Income Households For a Study of Food Acquisition
*John W Hall, Mathematica Policy Research
Keywords: Low Income Households, Address Based Sampling, Composite Size Measures, Food Acquisition, SNAP
The National Household Food Acquisition and Purchase Survey is collecting data from households of all income levels, but has targets defined by income and receipt of U.S. Supplemental Nutrition Assistance Program (SNAP) benefits. These targets require oversampling of certain hard to reach groups, particularly low-income households that do not receive SNAP. A pilot study has been completed that collected data on approximately 400 households in two sites and a national study will be fielded in early 2012 that will collect data from 5,000 households in 400 SSUs (50 PSUs) . The sample uses a three-stage design, where primary and secondary selection units (PSUs and SSUs) are selected using composite measures of size that reflect the projected prevalence of and sampling rate for each of four target groups: SNAP households and three income-defined strata of households not receiving SNAP. Within SSUs the study employs a dual-frame approach: administrative records to identify SNAP households and Address Based Samples (ABS) to identify others. The ABS frames are used as a replacement for traditional household listing. Field test results indicate that the sampling methods used efficiently identify the various target groups, but that some field listing was required because the ABS frame did not include all dwelling units. The paper will address all phases of sampling: how the composite size measures were constructed and their effectiveness; how the two frames were used within SSUs; problems in using the ABS frame and how they were resolved; managing the release of sample to meet targets for the rarer groups while not collecting data for too many of the others.