Managing Sample Release in Social and Economic Surveys

• Activity Number: 512
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Frank Potter*+
• Companies: Mathematica Policy Research
• Keywords: Survey Management ; Sample Size ; Survey Experimentation

Abstract

Sample surveys are often used in the evaluation of social programs and require specific target minimum numbers of respondents to achieve statistical precision objectives for multiple subpopulations. To achieve the target number of completed interviews with the smallest sample size in each subpopulation, a carefully managed release plan is required for each study subpopulation and we often use random subsamples. The process uses the selection of an "augmented" sample (using implicit stratification) that is substantially larger than is expected to be required (using "pessimistic" response rates). The "augmented" sample is then randomly partitioned into small subsamples using an algorithm that preserves the implicit stratification. The initial sample release for data collection is then based on "optimistic" response rates. Periodic assessments are conducted to determine if additional sample releases are needed and, if so, in which study subpopulation(s). In this paper we will describe the algorithm and provide guidance on how to manage a sample. This random partitioning can also be used to assess experimental data collection procedures.