Abstract: The Statistics of Income (SOI) Division within the Internal Revenue Service curates and maintains an extensive repository of tax data, offering researchers a valuable resource for evaluating the impacts of tax policies and exploring diverse research inquiries, including the analysis of income inequality. While the confidential data remains accessible only to a limited number of government analysts and researchers, SOI provides an accessible public use file for external researchers and data practitioners. However, safeguarding this public use file has grown increasingly difficult to protect through traditional statistical data privacy methods, as the vast amount of personal information available in public and private databases combined with enormous computational power create unprecedented privacy risks.

This presentation will cover the collaborative efforts of SOI and researchers at the Urban Institute, who are actively developing a solution: the creation of synthetic data that represent the statistical properties of the administrative data without revealing any individual taxpayer information. Additionally, Urban is building a prototype validation server that empowers researchers to indirectly conduct statistical analyses on administrative tax data. Researchers can accomplish this by evaluating their analyses using synthetic data and subsequently submitting them to the validation server. The server then produces a modified output with added noise, all the while maintaining the confidentiality of taxpayer information. In essence, this talk will address the lessons learned, best practices, and challenges encountered in the process of safely expanding access to administrative tax data.

https://clairemckaybowen.com/