637 – The Use of Auxiliary Data in Frame Development, Coverage Assessment, and Field Data Collection
Selecting a Sample from a Changing Frame of Program Beneficiaries: An Application of Adaptive Sampling
Eric Grau
Mathematica Policy Research
Probability samples are usually selected from a fixed sample frame, which is a close facsimile of the target population. In those situations, the probability of selection can be easily quantified. The sample design becomes more complicated when the sample frame is not fixed, but changes over time. In a previous paper, we described a complex design for a sample of people with disabilities receiving Supplemental Security Income and/or Social Security Disability Insurance benefits who had successful work experience, in the National Beneficiary Survey sponsored by the Social Security Administration. In this paper, we describe issues with the sample design's implementation in the first round of sampling, and discuss how the design was adapted to adjust for the fact that the sample frame changed over time. We show how the entire sample frame was divided into seven segments that were revealed over time, and how the sample allocations changed as we moved through time. We revisit some of the assumptions made when designing the sample, and assess what changes would be required in the next round of data collection.