The 2020 Census Disclosure Avoidance System TopDown Algorithm

Abstract

The Census TopDown Algorithm (TDA) is a disclosure avoidance system using differential privacy for privacy-loss accounting. The algorithm ingests the final, edited version of the 2020 Census data and the final tabulation geographic definitions. The algorithm then creates noisy versions of key queries on the data, referred to as measurements, using zero-Concentrated Differential Privacy. Another key aspect of the TDA are invariants, statistics that the Census Bureau has determined, as matter of policy, to exclude from the privacy-loss accounting. The TDA post-processes the measurements together with the invariants to produce a Microdata Detail File (MDF) that contains one record for each person and one record for each housing unit enumerated in the 2020 Census. The MDF is passed to the 2020 Census tabulation system to produce the 2020 Census Redistricting Data (P.L. 94-171) Summary File. This paper describes the mathematics and testing of the TDA for this purpose.

Assessing Statistical Disclosure Risk for Differentially Private, Hierarchical Count Data, With Application to the 2020 U. S. Decennial Census

“…We illustrate this in Section S6 of the supplement. The population invariants include the total populations of each state, total number of housing units in each census block, and number of occupied group quarters of each type in each census block (Abowd et al (2022)). It is unclear how possession of these quantities would affect our disclosure risk measures, since they are not easily related to the counts an adversary considers in our methods.…”

“…To describe the DP methods the Census Bureau is using for the 2020 decennial census, we draw from Abowd et al (2022). We focus on the release of the PL 94-171 file, which comprises 2020 census data used for redistricting.…”

“…The second column displays the total proportion of ρ allocated to each geographic level, and the third column displays the proportion of the ρ at that level allocated to the GVHR query. Values taken from Section 8.2 of Abowd et al (2022) 2.3 DP in the 2020 Census Data Products 10 displayed in the third column of Table 2. The remainder of the allocations can be found in Section 8.2 of Abowd et al (2022).…”

“…Values taken from Section 8.2 of Abowd et al (2022) 2.3 DP in the 2020 Census Data Products 10 displayed in the third column of Table 2. The remainder of the allocations can be found in Section 8.2 of Abowd et al (2022). The GVHR query in each census block, for example, is allocated a ρ of 0.103 × 3,945/4,097 ≈ 0.099.…”

“…To protect individual respondents' privacy in the 2020 decennial census public release data products, the U.S. Census Bureau applies a method satisfying a variant of differential privacy (Abowd et al (2022)). The Census Bureau decided to use a differentially private method, which they call the TopDown algorithm (TDA), because they determined that the risks of reidentifications when using methods employed in previous censuses, namely data swapping, are too great (Abowd (2018)).…”