Differentially private algorithms for detailed race and ethnicity in the 2020 census

Summary:

Tumult Labs designed a novel differentially-private algorithm that the US Census Bureau is considering for use in the release the Detailed Demographic and Housing Characteristics (DHC) Race & Ethnicity tabulations as part of the 2020 Census. This is a challenging data release problem because the tabulations contain counts of demographic and housing characteristics of the entire population of the US, combined with detailed racial and tribal categories at varying levels of geography. We describe two differentially private algorithms, one satisfying so-called "pure" differential privacy and one satisfying a well-studied variant of differential privacy, called zero-concentrated differential privacy.