Accuracy Impacts of Differential Privacy in the 2020 US Census

This paper was the final project for CS 763, Security and Privacy in Data Science and is a joint work with Keaton Leppanen and Shiyu Yang.

Survey paper about the Census Bureau’s impact on differential privacy and the empirical results on accuracy. We find that multiple analyses have shown that accuracy for certain marginalized groups - such as Native American tribal populations - will be much worse. We also found that populations of populous areas, such as cities, tends to be rounded-down by the TopDown algorithm’s optimization process, whereas rural areas often see increases in reported population.