Assessing the 2020 Census Final Report (2023) / Chapter Skim
Currently Skimming:
Appendix F: Additional Detail and Reference on Data Products and Disclosure Avoidance
Appendix F: Additional Detail and Reference on Data Products and Disclosure Avoidance
Pages 475-492
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 475... ...
(2022) , and Special Issue 2 of the Harvard Data Science Review (2022)
|
From page 476... ...
If k multiple, independent queries are posed to the posted data, the overall privacy loss budget equals ϵ 1 + · · · + ϵ k , the sum of the individual privacy loss budgets. So, maintaining an overall privacy loss budget of ϵ ∗ requires individual budgets that average ϵ ∗/k , which can be very small, allowing very little information to "leak through." The approach is similar in spirit to a minimax statistical procedure, one that bounds the maximum risk for all states of nature.
|
From page 477... ...
Realities include a wide variety of geographic and demographic domains that require different protection versus information trade-offs, a very large and to a degree unknowable set of future queries on the posted data, and facevalidity constraints. The Census Bureau's approach for the 2020 Redistricting and Demographic and Housing Characteristics (DHC)
|
From page 478... ...
Table F.1 provides details of the content, geographies, and timing for 2010 and 2020 Census data products, including the Redistricting File, Summary File 1/DHC File, Summary File 2/Detailed Demographic and Housing Characteristics (DDHC) Files A and B, Demographic Profile, and Public Use Microdata Sample.
|
From page 479... ...
for persons, Similar content to 2010 households, families, housing units by age, race, ethnicity, Block-level: Substantial deletions/moves to higher-level sex, household relationship, household type, housing geography (e.g., 11 imputation rate tables deleted; 6 occupancy/tenure, GQ, item imputation; cross-tabulations, household/family tables deleted; 6 household/family including sex by age, average household size by age; tables moved to census tracts [often without race/ethnicity iterations of selected tables by race/ethnicity of person iterations] ; 8 household/family/occupied housing join or household/family head tables moved to S-DHC [6 for census tracts, 2 for block groups without iteration]
|
From page 480... ...
(e.g., average household size by age, tenure) Geographies Block as lowest level State as lowest level (a major change announced May 31, 2023 -- original plans were to provide some tables for block groups, census tracts, places, and AIANNH areas)
|
From page 481... ...
Tables repeated for around 300 detailed race/ethnicity groups and around 1,200 AIANNH tribes/villages Geographies Census tract as lowest level: 61 total tables = 47 person DDHC-A and DDHC-B -- Census tract, plus place, county, tables, 14 housing tables state, AIANNH areas, nation; 4 total tables County as lowest level: 10 total tables (GQ) Release Schedule SF2: December 2011–April 2012 DDHC-A -- September 21, 2023 AIANSF: December 2012 DDHC-B -- Scheduled for September 2024 481
|
From page 482... ...
For "Block (or census tract) as lowest level," all higher-level geographies are included (e.g., block group, census tract, incorporated place, minor civil division, school district, county, AIANNH area, state)
|
From page 483... ...
, ϵ = 2 Census Bureau (2019) and associated summary metrics Product Baseline (housing tables)
|
From page 484... ...
discrete Gaussian mechanism to census/decade/2020/planning-management/process/ reduce outliers; use accuracy target for largest racial-ethnic disclosure-avoidance/2020-das-development.html group in any geography < 500 people of ±5 percentage points of enumerated value 95% of the time; optimize to bring off-spine geographies closer to the spine; separate post-processing for group quarters at the block group level June 2021: 2010 TDA used for redistricting tables; ϵ = 17 (person tables) , Fact sheet and associated summary metrics available under Demonstration ϵ = 2.5 (housing tables)
|
From page 485... ...
groups/areas with less accuracy in DHC demo product Demonstration nationalacademies.org/event/06-21-2022/2020-census- (e.g., AIAN population, Liebler; renters, Reynolds and Product Update #5 data-products-workshop-on-the-demographic-and-housing- Vink; rural population , Mueller; denominators for local characteristics-files age-adjusted death/disease rates, Werner; small Traffic Analysis Zones, Kaneff) August 2022: 2010 TDA used for DHC tables; combined DHC ϵ = 46; See https://www2.census.gov/programs-surveys/decennial/ Demonstration improvements: changed lowest level of geography for 2020/program-management/data-product-planning/2010 Product Update #6/ some tables from state/county to census tract; iterated demonstration-data-products/02-Demographic_and_ DHC Production tables for sex by single year of age at census tract level; Housing_Characteristics/2022-08-25_Summary_File/2022 Settings greater accuracy for some tables but not group quarters 08-25_Factsheet.pdf; Census Data Stewardship Executive people by age; limited reduction in person-housing unit Policy Committee decided final PLB parameters for DHC, inconsistencies November 2022 (made public in April 2023; see below)
|
From page 486... ...
NOTES: TDA, Top Down Algorithm; DHC, Demographic and Housing Characteristics (File) ; DDHC, Detailed DHC; CNSTAT, Committee on National Statistics of the National Academies of Sciences, Engineering, and Medicine; PLB: privacy loss budget; Updates #2–4 released as privacy protected microdata files (PPMFs)
|
From page 487... ...
About 20% of the Nevada governmental units that are eligible for state revenue sharing had fewer than about 400 people in 20206 and thus would likely have highly variable population estimates from the housing unit-based method using data from the 2020 Redistricting File. More accurate estimates of persons per occupied housing units are now available in the DHC File released in May 2023.7 F.3.2 Public Health Estimates Public health analysts require small-area census data for planning, implementation, and evaluation of public health practices.
|
From page 488... ...
. For age-adjusted estimates of hospitalizations and emergency department visits for asthma, county data were reasonably comparable between the 2010 DHC Demonstration File and the original 2010 SF1, but much less so for census tracts.
|
From page 489... ...
F.3.4 Special Areas for Local Planning State and local governments and regional planning organizations often aggregate census blocks, block groups, or census tracts to form their own areas for planning purposes. Kaneff (2022)
|
From page 490... ...
compared the March 2022 version of the 2010 DHC Demonstration File with the original 2010 SF1 and found that the two data sets differed more for households in rental-majority areas (census tracts) compared with owner-majority areas and that the differences were particularly pronounced for households with children and large households in rentalmajority areas.
|
From page 491... ...
APPENDIX F 491 the demonstration and original datasets for most race and ethnic groups, with the exceptions of Asian people and people who were not Hispanic.
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.