The 2020 Census Disclosure Avoidance System TopDown Algorithm

Background: The 2020 US Census will use a novel approach to disclosure avoidance to protect respondents’ data, called TopDown. This TopDown algorithm was applied to the 2018 end-to-end (E2E) test of the decennial census. The computer code used for this test as well as accompanying exposition has recently been released publicly by the Census Bureau. Methods: We used the available code and data to better understand the error introduced by the E2E disclosure avoidance system when Census Bureau applied it to 1940 census data and we developed an empirical measure of privacy loss to compare the error and privacy of the new approach to that of a (non-differentially private) simple-random-sampling approach to protecting privacy. Results: We found that the empirical privacy loss of TopDown is substantially smaller than the theoretical guarantee for all privacy loss budgets we examined. When run on the 1940 census data, TopDown with a privacy budget of 1.0 was similar in error and privacy loss to that of a simple random sample of 50% of the US population. When run with a privacy budget of 4.0, it was similar in error and privacy loss of a 90% sample. Conclusions: This work fits into the beginning of a discussion on how to best balance privacy and accuracy in decennial census data collection, and there is a need for continued discussion.

Through the lens of the system of differential privacy specifications developed in Part I of a trio of articles, this second paper examines two statistical disclosure control (SDC) methods for the United States Decennial Census: the Permutation Swapping Algorithm (PSA), which is similar to the 2010 Census’s disclosure avoidance system (DAS), and the TopDown Algorithm (TDA), which was used in the 2020 DAS. To varying degrees, both methods leave unaltered some statistics of the confidential data—which are called the method’s invariants—and hence neither can be readily reconciled with differential privacy (DP), at least as it was originally conceived. Nevertheless, we establish that the PSA satisfies -DP subject to the invariants it necessarily induces, thereby showing that this traditional SDC method can in fact still be understood within our more-general system of DP specifications. By a similar modification to -zero concentrated DP, we also provide a DP specification for the TDA. Finally, as a point of comparison, we consider the counterfactual scenario in which the PSA was adopted for the 2020 Census, resulting in a reduction in the nominal privacy loss, but at the cost of releasing many more invariants. Therefore, while our results explicate the mathematical guarantees of SDC provided by the PSA, the TDA and the 2020 DAS in general, care must be taken in their translation to actual privacy protection—just as is the case for any DP deployment.

Differential Privacy in the 2020 Decennial Census and the Implications for Available Data Products

The Census Bureau has announced new methods for disclosure control in public use data products. The new approach, known as differential privacy, represents a radical departure from current practice. In its pure form, differential privacy techniques may make the release of useful microdata impossible and limit the utility of tabular small-area data. Adoption of differential privacy will have far-reaching consequences for research. It is likely that scientists, planners, and the public will lose the free access we have enjoyed for six decades to reliable public Census Bureau data describing US social and economic change.

Census statistics play a key role in public policy decisions and social science research. However, given the risk of revealing individual information, many statistical agencies are considering disclosure control methods based on differential privacy, which add noise to tabulated data. Unlike other applications of differential privacy, however, census statistics must be postprocessed after noise injection to be usable. We study the impact of the U.S. Census Bureau’s latest disclosure avoidance system (DAS) on a major application of census statistics, the redrawing of electoral districts. We find that the DAS systematically undercounts the population in mixed-race and mixed-partisan precincts, yielding unpredictable racial and partisan biases. While the DAS leads to a likely violation of the “One Person, One Vote” standard as currently interpreted, it does not prevent accurate predictions of an individual’s race and ethnicity. Our findings underscore the difficulty of balancing accuracy and respondent privacy in the Census.

Variation in Incidence of ESKD Among Individuals of Native Hawaiian/Pacific Islander Race Based on Data From the US Renal Data System

Emerging discussions on the responsible government use of algorithmic technologies propose transparency and public participation as key mechanisms for preserving accountability and trust. But in practice, the adoption and use of any technology shifts the social, organizational, and political context in which it is embedded. Therefore translating transparency and participation efforts into meaningful, effective accountability must take into account these shifts. We adopt two theoretical frames, Mulligan and Nissenbaum’s handoff model and Star and Griesemer’s boundary objects, to reveal such shifts during the U.S. Census Bureau’s adoption of differential privacy (DP) in its updated disclosure avoidance system (DAS) for the 2020 census. This update preserved (and arguably strengthened) the confidentiality protections that the Bureau is mandated to uphold, and the Bureau engaged in a range of activities to facilitate public understanding of and participation in the system design process. Using publicly available documents concerning the Census’ implementation of DP, this case study seeks to expand our understanding of how technical shifts implicate values, how such shifts can afford (or fail to afford) greater transparency and participation in system design, and the importance of localized expertise throughout. We present three lessons from this case study toward grounding understandings of algorithmic transparency and participation: (1) efforts towards transparency and participation in algorithmic governance must center values and policy decisions, not just technical design decisions; (2) the handoff model is a useful tool for revealing how such values may be cloaked beneath technical decisions; and (3) boundary objects alone cannot bridge distant communities without trusted experts traveling alongside to broker their adoption.

INTRODUCTION: The American College of Gastroenterology advocates that colorectal cancer screening should start earlier in African Americans because of the higher incidence and more aggressive nature in this population. We aimed to address colonoscopy screening disparities among the different races and genders investigating valuable significant contributors. METHODS: We reviewed GIQuIC data from 3 hospitals and 5 endoscopy outpatient surgical centers from October 2018 to December 2019. An analysis was created utilizing Highmark Health claims and the US Census Bureau American Community Survey data showing the strongest correlation for a lack of colorectal cancer screening by race and lower socioeconomic status (Figure 1). Through utilization of this data and GIQuIC data, we analyzed race and age of colorectal cancer screenings among African American and Caucasians. RESULTS: Among all screening colonoscopies available in the GIQuIC data, 84.1% constituted whites and 11.5% constituted African Americans (P < 0.0001) which is consistent with the low rate of African American screenings in the census data. Of all screenings, 54.7% were females and 45.3% males (P < 0.0001). We further looked at the data by race and gender (Table 1) and screening rates by age group (Table 2). Whites had screenings at age 50 more often than age appropriate screening for African Americans (age 45). Interestingly, colonoscopy prep quality was the lowest in black males, while it was the highest for white females. CONCLUSION: We found that African Americans are less likely than Caucasians to have screening tests for colorectal cancer in our region. This is consistent with previous studies. This may be due to lower socioeconomic status, lack of access in underserved areas, and lack of awareness/education. Better resources are required to educate the African American population regarding timely colonoscopy screening at age 45, colonoscopy prep instruction, and identifying barriers to access as key indicators to improve screening overall. We plan to target the underserved areas in the near future to improve CRC screening in the African American population. There was also an overwhelmingly higher number of females screened when compared to males. Is it because males are more apprehensive about colonoscopy? Further studies are needed to investigate such findings.Figure 1.: US Census Bureau American Community Survey data.Table 1.: Race and Gender.Table 2.: Screening Rates by Age Range.

The article presents a model developed to estimate the undercount stemming from lowered response among sub-populations of 1 st and 2 nd generation Latino immigrants if a question on citizenship is included in Census 2020. The analysis is relevant to census efforts wherever socioeconomic and sociopolitical disparities result in differential census participation. The model is referred to as a “cascade” model because it examines successive causes of undercount in the course of non-response follow-up (NRFU), partial household omission in “complex” households, and omission of low-visibility housing units from the U.S. Census Bureau’s Master Address File (MAF). The analysis also examines the likelihood of enumeration errors from the U.S. Census Bureau’s proposed reliance on administrative records for enumerating non-responding housing units. The model incorporates data from an 8-county survey of Latino immigrants regarding their willingness to participate in Census 2020 if it includes a question on citizenship. It shows that systematic differences in the size of responding and non-responding households will undermine reliability of hot-deck imputation. The conclusion is that adoption of inadequately-tested “modernized” census procedures exacerbates differential undercount of immigrant populations and contributes significantly to geographic disparities in the census count and erodes the reliability demographic profile of areas with higher-than-average concentrations of immigrants.

The Census Bureau has recently released a plan to ensure that all statistics released as part of the 2020 Census are protected by a noise injection algorithm that satisfies the definition of differential privacy. Using 1940 Census data that has been treated with a noise injection process similar to that proposed for the 2020 Census, we explore the utility of the 1940 data by analyzing how the additional noise might affect standard uses of decennial census data. We consider three separate uses of decennial census data: oversampling populations in surveys, screening operations for surveys of rare populations, and allocating federal funds to specific areas. We find that for use cases that involve large populations, the effects of noise injection are relatively modest. Nonetheless, the noise injection can lead to sampling-frame coverage issues for surveys of rare populations and to substantial misallocations of funds to local areas.

Evaluating Age, Sex, Racial, and Ethnic Representation in Acute Ischemic Stroke Trials, 2010 to 2020: A Systematic Review and Meta-Analysis.

As this article is published, the U.S. Census Bureau is completing work for the twenty third decennial census of the United States. Once again, the MAF/TIGER system served as the geospatial infrastructure supporting numerous census operations and data collection, tabulation, and dissemination activities. From data collection to data dissemination we trace the recent activities of the 2010 Decennial Census of the United States to illustrate the role maps and geospatial data play in an increasing variety of public and private sector activities across the nation. To ensure a successful 2010 Census, millions of maps had to be created. This article will give an overview of the automated mapping system designed to create these maps. This includes a discussion about associated software needed and the variety of map types that were developed. Finally, future map production and geospatial activities at the Census Bureau will be discussed.

The dissemination of synthetic data can be an effective means of making information from sensitive data publicly available with a reduced risk of disclosure. While mechanisms exist for synthesizing data that satisfy formal privacy guarantees, these mechanisms do not typically resemble the models an end-user might use to analyse the data. More recently, the use of methods from the disease mapping literature has been proposed to generate spatially referenced synthetic data with high utility but without formal privacy guarantees. The objective for this paper is to help bridge the gap between the disease mapping and the differential privacy literatures. In particular, we generalize an approach for generating differentially private synthetic data currently used by the US Census Bureau to the case of Poisson-distributed count data in a way that accommodates heterogeneity in population sizes and allows for the infusion of prior information regarding the underlying event rates. Following a pair of small simulation studies, we illustrate the utility of the synthetic data produced by this approach using publicly available, county-level heart disease-related death counts. This study demonstrates the benefits of the proposed approach’s flexibility with respect to heterogeneity in population sizes and event rates while motivating further research to improve its utility.

American trypanosomiasis or Chagas disease caused by Trypanosoma cruzi affects many mammals, including humans and dogs, in all Latin American countries outside the Caribbean and increasingly also in the southern United States. Dogs are considered as reliable sentinels and have been identified as an important risk factor for the disease in humans in endemic countries. Factors that determine American trypanosomiasis in dogs may therefore have public health relevance. Associations of different environmental, locational, and pet owner socioeconomic conditions were evaluated retrospectively as potential risk factors for American trypanosomiasis status in dogs in a case-control study. Laboratory-confirmed cases received at the Texas A&M University Veterinary Medical Teaching Hospital between the years 2000 and 2012 and candidate risk factor variables extracted from publicly available environmental data and 2010 US Census Bureau were used. The sample included 42 dogs serologically positive and 82 dogs serologically negative determined by indirect immunofluorescent assay. The diagnostic titer was 1:160 (case). Univariate logistic regressions followed by stepwise multivariate logistic modeling were used for variable screening and to determine the strengths of variable associations with case status. Total Edge Contrast Index (odds ratio [OR] = 3.35, 95% confidence interval [CI] 3.10, 3.62), residing in homes that had rural addresses (OR = 2.48, 95% CI 2.43, 2.53), total number of owner occupied housing units in a neighborhood with a householder who is Hispanic or Latino (OR = 1.66, 95% CI 1.04, 2.66), and the total number of housing units in a neighborhood that were built on or prior to year 1980 (OR = 2.22, 95% CI 1.94, 2.55) were identified as risk factors. Suitable awareness campaigns and future research that considers pet owner housing and socioeconomic circumstances are necessary for effective prevention and control of this disease among dogs.

In order to determine the relationship between socioeconomic deprivation and nonalcoholic fatty liver disease (NAFLD)/nonalcoholic steatohepatitis (NASH), we retrospectively reviewed the electronic medical records of 1,430 patients in a large tertiary health care network in New York. These patients underwent liver biopsy over a 10‐year period and were included in our study if they had evidence of NAFLD/NASH on liver biopsy. Zip codes were used to obtain data necessary to derive the social deprivation index (SDI) from the US Bureau of the Census. The high‐SDI group was compared to the low‐SDI group. Univariate and multivariate logistic regressions were performed to assess association between socioeconomic factors and NAFLD parameters, including presence of NASH (NAFLD activity score >4), moderate to severe steatosis (>33%), and significant fibrosis (S2‐S4). We included 614 patients with NAFLD/NASH; the median SDI was 31.5. Hemoglobin A1c values were higher in the high‐SDI group compared to the low‐SDI group (6.46 vs. 6.12, P = 0.02). Socioeconomic factors, such as private versus public health care, percentage being foreign born, percentage without a car, percentage with higher needs (<5 years old and >65 years old), and percentage currently living in renter‐occupied and crowded housing units, showed statistically significant associations in predicting NASH. After adjusting for patient age, sex, race, body mass index, and diabetes, we saw a significant association between four or more socioeconomic parameters in predicting NASH (odds ratio [OR], 1.71; 95% confidence interval [CI], 1.099‐2.856; P = 0.0190) and six or more socioeconomic parameters in predicting severe steatosis (OR, 1.498; 95% CI, 1.031‐2.176; P = 0.0338) but no significant correlation between the number of socioeconomic parameters and significant fibrosis. Conclusion: Greater number of socioeconomic determinants (four or more) are associated with greater severity of NASH. Awareness of NAFLD/NASH needs to be raised in communities with high socioeconomic deprivation.