Privacy in Census Data - Math Meets Policy

Explore the intersection of mathematics and policy in this lecture on privacy in Census data. Delve into the Census's "differentially private" disclosure avoidance mechanism for the 2010 Decennial release, which intentionally introduces controlled random noise to released numbers. Examine the TopDown algorithm's mathematical aspects, focusing on its implications for redistricting. Learn about the contrasting reactions from different sectors, including computer science departments and community organizers. Investigate the algorithm's impact on redistricting, the Voting Rights Act of 1965, and racial polarization statistics. Analyze a case study from Irving, TX, and explore reconstruction algorithms, error variance, and budget allocation in the Census code. Gain insights into the challenges and future directions of balancing data privacy with accurate representation in Census data.

Intro
ABSTRACT
DIFFERENTIAL PRIVACY
CENSUS HIERARCHY
CENSUS TOPDOWN
REDISTRICTING AND THE PL94-171 RELEASE
THE VOTING RIGHTS ACT OF 1965
STATS AND RACIAL POLARIZATION
LEGAL FICTIONS AND STANDARD MESSY DATA
CASE STUDY: IRVING, TX
RECONSTRUCTION ALGORITHM
TOYDOWN: BASIC TREES
RUNNING ACTUAL CENSUS CODE
ERROR VARIANCE IN TOYDOWN
IDEAL BUDGET ALLOCATION
OFF-SPINE ERROR AND "HIERARCHICAL INTEGRITY"
WHAT'S NEXT?