Synthetic Data and Social Science Research: Accuracy Assessments and Practical Considerations from the SIPP Synthetic Beta

Synthetic microdata - data retaining the structure of original microdata while replacing original values with modeled values for the sake of privacy - presents an opportunity to increase access to useful microdata for data users while meeting the privacy and confidentiality requirements for data providers. Synthetic data could be sufficient for many purposes, but lingering accuracy concerns could be addressed with a validation system through which the data providers run the external researcher's code on the internal data and share cleared output with the researcher. The U.S. Census Bureau has experience running such systems. In this chapter, we first describe the role of synthetic data within a tiered data access system and the importance of synthetic data accuracy in achieving a viable synthetic data product. Next, we review results from a recent set of empirical analyses we conducted to assess accuracy in the Survey of Income & Program Participation (SIPP) Synthetic Beta (SSB), a Census Bureau product that made linked survey-administrative data publicly available. Given this analysis and our experience working on the SSB project, we conclude with thoughts and questions regarding future implementations of synthetic data with validation.