What We Know So Far about 2020 Census Data Quality

By Meeta Anand

On April 26, 2021, the U.S. Census Bureau released apportionment data, including state population totals that are the basis for allocating seats in the U.S. House of Representatives and the number of seats awarded to each state for the next decade. It also released national and state resident population counts and two sets of operational metrics, which provide some early insights into census data quality.

While we are only at the tip of the iceberg for 2020 Census data releases, here is what we do know so far:

The number of people who lived in the United States on April 1, 2020. The 2020 Census resident population was 331,449,281, which includes all people, regardless of citizenship, living in the 50 states and the District of Columbia as of April 1, 2020. This represents a 7.4 percent increase from the 2010 Census. This population total includes U.S. troops deployed overseas but stationed in the United States during the census.

The number of congressional seats per state. The apportionment process determines the number of seats in the House of Representatives to which each state is entitled for the next decade, reflecting relative shifts in population between the states over the past decade. The number of seats in the House of Representatives is capped by law at 435. Following the 2020 Census, 13 states either gained or lost congressional seats. The apportionment results will take effect starting in the 118th Congress in 2023. This map shows the history of congressional apportionment following each decennial census from 1910 through 2020.

The apportionment population per state. The apportionment population totals were also released, which are different from the resident population totals. They include the federally affiliated overseas population and exclude the District of Columbia.* The apportionment population numbers are used to determine the allocation of seats in the House using the Method of Equal Proportions.

The way that data on households and group quarters were collected. On April 26, 2021, the Census Bureau also released its first set of operational metrics for the 2020 Census for the nation, 50 states, the District of Columbia, and Puerto Rico. These quality metrics tell us how the bureau accounted for each address on its master list, including group facilities (such as college dorms) by occupancy status (occupied, vacant, or deleted as nonexistent). For housing units, those methods included self-response, in-person interview with a household member, proxy interview (such as with a neighbor or landlord), administrative records, and count imputation.

A breakdown of how addresses were resolved by household size and use of the Census ID. On May 28, 2021, the Census Bureau released a second set of operational metrics with a focus on vacant and occupied housing units and the household size of occupied units. The metrics for single- and two-person households show the method of enumeration (percentage resolved by self-response, in-person interview with a household member, proxy interview, administrative records, and imputation). They also include the percentage of households that responded to the 2020 Census online without a Census ID (referred to as non-ID processing), for the nation, the 50 states, and the District of Columbia. Nationwide, 8 percent of households that responded online did so without the Census ID assigned to their address.

Benchmarks the Census Bureau uses to evaluate census accuracy. Some of the many measures the Census Bureau uses to evaluate the accuracy of census results are now available for the 2020 Census.

  • Demographic Analysis. The Census Bureau released the 2020 Demographic Analysis estimates on December 15, 2020. Demographic Analysis (DA) is one of two methods the Census Bureau uses to estimate net under- and overcounts in the census (the second being the Post-Enumeration Survey, discussed below). The bureau produced three DA estimates of the population: low (330,730,000), middle (332,601,000), and high (335,514,000). The different estimates reflect different underlying assumptions, primarily related to immigration. The 2020 Census national population total was 0.3 percent below the middle DA estimate. The DA also provides population statistics by age, gender, and selected race and ethnicity categories. Future releases by the Census Bureau should allow for comparisons of statistics for racial groups.
  • Population Estimates. The Census Bureau released the 2020 national and state population estimates on December 22, 2020. The Census Bureau’s Population Estimates Program is a partnership between states and the Census Bureau that has produced estimates annually for more than four decades. These estimates are produced for states, counties, and places of 50,000 or greater annually, and for all governmental units every other year, in between decennial censuses. The percentage difference between 2020 Census state population totals and state population estimates ranged from being 4.5 percent above the estimate (New Jersey) to 3.3 percent below (Arizona). The estimates in this series are produced from a baseline of decennial census data. For example, the 2020 population estimates used 2010 Census numbers as a base and did not include any data collected as part of the 2020 Census. For the next 10 years, starting in 2021, the population estimates will use 2020 Census data as the base. More information on population estimates and an initial analysis of the 2020 numbers can be found here.
  • Post-Enumeration Survey. The Post-Enumeration Survey (PES) measures census accuracy by collecting data about April 1, 2020 household composition from a representative sample of housing units nationwide once census door-knocking is complete in each area. The responses are compared to the information collected in the census from those homes; the findings are then projected onto all similar households, block by census block, to produce measurements of how well the census counted various population subgroups, including by age (cohorts), gender, race, Hispanic origin, and housing tenure (owner or renter). The PES provides estimates of net over- and undercounts, as well as gross errors such as omissions (people missed) and duplications (people counted twice) in the decennial census. PES data are expected to be released sometime in 2022.

The ways that third parties will assess the quality of the 2020 Census. The Census Bureau has engaged three independent groups to evaluate the quality of the 2020 Census. JASON is an independent group of scientists and engineers that advises the U.S. government on matters of science and technology. The JASON report recommended communication strategies, supported the release of operational metrics, and encouraged drawing from experiences of the 2020 Census to inform 2030 Census planning. The American Statistical Association (ASA) released a report in October 2020 proposing various measures of quality, accuracy, and coverage to be applied to 2020 Census data. Experts selected by ASA continue to work with the Census Bureau to examine 2020 Census quality indicators; their findings will inform two reports about 2020 Census quality that ASA expects to release this summer and fall. Finally, at the Census Bureau’s request, the Committee on National Statistics of the National Academy of Sciences has assembled a panel of experts to review 2020 Census data quality and make recommendations for further research, including for the 2030 Census.

The expected timing of the release of neighborhood, place, and demographic data. Barring court orders or unforeseen circumstances, the Census Bureau will publish data for the smallest geographic areas (census blocks) by voting age, race, and ethnicity in August and September 2021. The Census Bureau has committed to releasing this information — used for congressional, state legislative, and local redistricting — in so-called “legacy format” by August 16 and in easier-to-use data files by September 30.  Additional operational metrics will be released during this time frame as well. Coupled with the redistricting data, these metrics will support further analysis of the quality of 2020 Census data, as well as a deeper understanding of how individual neighborhoods and communities were counted.

These initial census counts, estimates, and metrics allow stakeholders to begin evaluating their own work as well as that of the Census Bureau. Soon, the Census Bureau will release additional data that will be used for redistricting purposes and will help inform how federal funds will flow to communities. The Census Bureau will continue to release 2020 Census data products over the next several years, including the Demographic Profiles and Demographic and Housing Characteristic files. The American Community Survey — the evergreen part of the census — produces a wide range of data about the socio-economic characteristics of our communities on an annual basis, which helps local officials, community leaders, and businesses track and respond to changing community conditions and profiles.


Meeta Anand is the census consultant to The Leadership Conference Education Fund.

 

*The federally affiliated overseas population includes U.S. military members and federal civilian employees stationed overseas, and any dependents living with them. They are counted at their “home of record” based on administrative data from the Department of Defense and federal agencies.