Optimizing Survey Cost-Error Tradeoffs: A Multiple Imputation Strategy Using the Census Planning Database

• Activity Number: 593
• Type: Topic Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Government Statistics Section
• Author(s): Shin-Jung Lee*
• Companies: University of Michigan
• Keywords: Census Planning Database ; Survey design ; Nonresponse ; ACS ; CPS ; Imputation

Abstract

The concerns on survey errors and the rise in survey costs are exacerbated by the increase in survey nonresponse. To optimize the cost-error tradeoffs we propose, in a multi-phase survey, to enhance the representativeness of respondents and the prediction of missing responses through adaptive sampling and imputation. The adaptive sampling continuously benchmarks a focal survey to a high quality benchmark survey. The nonresponse imputation builds upon the Census Planning Database (PDB). Models predicting sample characteristics are fitted to the data from the benchmark survey, which shares the same contextual information (PDB) as the focal survey. The PDB bridges the prediction of missing data with the data from respondents and the benchmark. With better prediction of nonresponse, estimation of nonresponse errors and decision on nonresponse follow-up become a data-driven strategy. The 2010 PDB, ACS and CPS are used to illustrate the proposed method. The method is evaluated by various cost and error decisions, including comparison to the conventional fixed sampling design that uses post-survey weighting and nonresponse follow-up.