![IconGems-Print](images/IconGems-Print.png)
591 – Synthetic Data and Data Disclosure
Preserving Privacy in Person-Level Data for the American Community Survey
Rolando RodrÃguez
U.S. Census Bureau
Michael H. Freiman
U.S. Census Bureau
Jerome P. Reiter
Duke University
Amy D. Lauger
U.S. Census Bureau
The Census Bureau is researching model-based synthetic person-level data that maintain many of the properties of the original American Community Survey (ACS) data while protecting individual privacy. Protecting the ACS while maintaining data quality presents particular challenges because of the ACS's sample weighting, the survey's large number of variables and the small geographies for which ACS data are desired. This paper discusses the reasons adapting existing differential privacy methods is difficult and describes the approaches we are investigating to protect the data, including tree-based methods for categorical or discrete variables and regression for continuous variables.