Conducted every 10 years, the Census is a comprehensive count of the people living in the United States and its territories. This information is used to determine the number of seats in the House of Representatives apportioned to each state, to draw legislative districts, and to direct federal funding each year.
The 2020 Census encountered delays and incomplete data collection due to the COVID-19 pandemic and the Trump administration’s mismanagement. This analysis lays out the challenges and shortcomings of the 2020 Census, and discusses practices for responsible, science-based use of the challenging Census data for redistricting to ensure fair representation of historically undercounted groups, such as immigrants, low-income communities, and people of color, and to guard against gerrymandering.
The 2020 Census Delays, Changes, and Complications for Accuracy and Redistricting
This is a condensed, online version of the report. Access to all figures and full report are available through download of the PDF.
Understanding the 2020 Census
The United States decennial census is the “once-a-decade population and housing count of all 50 states, the District of Columbia, Puerto Rico, and the Island Areas as required by the US Constitution” (USCB 2017). It is performed by the US Census Bureau, the nation’s largest statistical agency, which seeks to provide timely and quality data to the US government about its people. Census results are used to determine the number of seats in the House of Representatives apportioned to each state, to draw congressional and state legislative districts, and to direct over $1.5 trillion in federal funding each year.
The Census Bureau ran into many problems conducting the decennial census in 2020, beginning with the COVID-19 pandemic. This public health crisis dramatically affected census deadlines, in part because of concern for the health and safety of the bureau’s field workers, who numbered more than 500,000. In March 2020, the Census Bureau suspended all field operations and requested deadline extensions (USCB 2020a). The federally mandated deadline for delivering apportionment counts is December 31, while March 31 of the following year is the deadline for providing a redistricting data file to allow redrawing of legislative districts. The Census Bureau requested an updated deadline of April 30, 2021, for apportionment counts and July 31, 2021, for redistricting data.
These schedule changes were recommended by the Census Bureau to ensure high-quality data collection—tabulating the census data on a truncated timeline would likely introduce shortcuts to tabulating the data, which would likely undercount people in regions hardest hit by the pandemic or by severe weather in 2020 as well as historically underrepresented populations. Once proper safety precautions were in place, regular field operations resumed in the summer of 2020. However, on August 3, 2020, the Census Bureau announced that instead of meeting the rescheduled deadlines, it would revert back to the federally mandated deadlines, ending field collection a month earlier than expected on October 15, 2020 (USCB 2020b).
This announcement followed on the heels of multiple changes to the census timelines. The initial extensions were approved by President Trump and Department of Commerce Secretary Wilbur Ross. They were introduced in the House in May and the Senate in June, while the Trump administration added two new political appointees to the Census Bureau (Wang 2020a). However, internal pressure at the Bureau from Secretary Ross to accelerate census operations resulted in the counting efforts being cut short in September (Fontenot 2020). This change, as noted internally by Bureau employees, raised the possibility that “serious errors discovered in the data may not be fixed—due to lack of time to research and understand the root cause” (Wang 2020b). The deadline extensions affecting the release of apportionment counts and redistricting data caused confusion and problems in a number of states. On April 26, 2021, the Bureau released apportionment counts of state congressional seats (Adams 2021). It then released a schedule indicating that it would release granular redistricting population data, in an older “legacy” format, in mid-to-late August, with the final redistricting data delivery expected by September 30, 2021 (USCB 2021; Whitehorne 2021). Some states have state constitutional deadlines to draw new districts based on when they expected to receive data according to the old federal schedule. Coupled with the aforementioned concerns about the quality of the 2020 Census data, this schedule mismatch compounds issues at the state level. Some states may implement inaccurate maps as a result, because they will use existing (rather than redrawn) maps or need state government action to adjust deadlines to avoid courts drawing maps (Cohn 2020; Rudensky, Li, and Limon 2021). The uproar concerning a citizenship question further compounded the problems plaguing the census. In March 2018, the Census Bureau announced its plan to add a question asking respondents about their US citizenship in order to help enforce the Voting Rights Act and its protections against voting discrimination (USCB 2018). The introduction of this question was met with opposition. Complaints included that the question lacked rigorous testing before its planned addition to the census and that it risked undercounting historically underrepresented groups across the country (Ciurczak and McEnroe 2018; Jarosz 2018). The proposed question led to multiple lawsuits (Wang 2018). Ultimately, a federal judge signed a court order to block the citizenship question from making its way onto the 2020 census. However, this turmoil during the years leading up to, and during, the taking of the 2020 census dramatically affected the public’s trust in and perception of the census. The federal government consequently felt the need to spend millions of dollars on advertising campaigns intended to convince the population that responding to the census was safe (King 2020). Concerns about these issues lead experts to worry about potential undercounting of various groups in the country and how undercounts might affect federal funding allocation, redistricting efforts, and representation. These worries include potential undercounting of Hispanic populations in many communities and of young children, as well as underestimates of population at the state level (Epstein and Medina 2021; O’Hare et al. 2020; Bahrampour, Rabinowitz, and Mellnik 2021). In this report, we seek to identify communities where undercounting and redistricting may affect representation, and we discuss best practices for censuses going forward to ensure communities are properly counted and represented.
To best understand the community-level impacts of these changes and concerns, we focused on a few communities across the United States where undercounts could have a significant impact on communities of interest, as these states are losing a congressional district, and to illustrate differences in the redistricting process. The first we highlight is Ohio, whose legislative districts are drawn by a legislative commission. We also worked with Michiganders, whose districts are going to be drawn by a citizen commission.
Beginning in 2021, Ohio adopted redistricting reforms to provide a better process to its citizens for drawing new districts (PGP n.d.). For congressional districts, the Ohio General Assembly draws the maps and must pass the redistricting plan in the legislature with bipartisan support. For state legislative districts, the Ohio Redistricting Commission, composed of the governor, state auditor, secretary of state, and one official appointed by each state legislative leader, draws the maps and the plan must pass the commission with bipartisan support. Each of these methods has multiple procedures in place to be used in the event of an impasse in order to ensure that adequate maps are drawn and enacted (Wendel 2021). Although an imperfect system still based on elected officials, the system improves on former laws in Ohio that enabled the state legislators to produce biased maps and entrench themselves in power (Thomas-Lundborg 2018). These changes were coupled with the addition of language to the Ohio state constitution requiring that the state legislative and congressional districts be contiguous, compact, and preserve existing communities. The deadlines to adopt a mapping plan are September 1 for the state legislative districts and September 30 for the congressional districts, dates soon approaching in a year in which census data is heavily delayed. Traditionally, the Census Bureau provides population data to states on April 1, providing plenty of time for contractors to draw maps that meet the September deadlines and required public hearings beforehand. With only legacy data files available by August 16, the pressure is on Ohio University, this year’s contractor, to analyze Ohio’s 2020 census results. This is the first time that the university is processing the data; the contractor in 2011 was Cleveland State University. There are concerns that the university’s commitment to turn the legacy data files into usable maps in the compressed deadline of two weeks is unrealistic. This concern couples with political disagreements in the state, which include pushes for a constitutional amendment to move deadlines and taking the case to extend deadlines to the Ohio Supreme Court (Kasler 2021). In short, the problems of the 2020 census demonstrate the impact delayed data can have on constitutionally mandated deadlines and how that delay may dictate the ultimate quality of the districts that are drawn in 2021. At the community level, advocates with the League of Women Voters of Ohio and Common Cause Ohio are concerned that compressed deadlines will result in poor representation of communities across Ohio (LWVO n.d.; CCO n.d.). These deadlines put the required public hearings on a much shorter schedule. Empirical evidence from census response rates also points to low-income areas of Ohio such as Youngstown and parts of Appalachia being potentially undercounted, which, coupled with the anticipated reshuffling of districts, could prove disastrous for these communities that rely heavily on government assistance (CUNY n.d.; Skolnick 2021). This possibility is also in play for two of Ohio’s poorest cities, Toledo and Cleveland, which look to be undercounted and are likely to be greatly affected by redistricting results based on the 2020 census. Clearly, the compressed census timeline will almost certainly impact the voices of Ohio communities, their elected representation over the next four or ten years, and the well-being of communities across the state.
In November 2018, Michigan voters passed a constitutional amendment to establish an Independent Citizens Redistricting Commission to draw lines for the state legislature and congressional seats. The commission includes four members who identify as Republican, four members who identify as Democrat, and five independent members who do not identify with a major party. This commission prioritizes contiguous districts that preserve communities of interest, do not advantage any parties or officials, reflect existing boundaries, and are compact (MICRC n.d.a). The commission aims to use the time period between the autumn of the census year (in this case, 2020) through the autumn of the redistricting year to meet, host town halls, and draw maps before the draft map deadline of September 17 (MICRC n.d.b). This process would involve 45 days of public comment before the maps were finalized by November 1 and then become law on December 31, subsequently taking effect for the next year’s elections. However, due to the delay in the release of 2020 census data to September 30, the commission and the Michigan secretary of state filed a lawsuit to extend the deadline, allowing for maps to be available on December 11 (Benson and ICRC 2021). This extension would allow maps to be approved on or before January 25, 2022, allowing the constitutionally mandated 45-day period of public comment. The lawsuit currently sits before the Michigan Supreme Court, with lawyers arguing against the case suggesting that the court should not intervene until the commission shows it truly cannot meet deadlines (Martinez-Beltran 2021). This delay, and general lack of clarity for Michiganders, could result in communities across the state lacking representation, in direct conflict with reforms passed in 2018. Community advocates are also increasingly concerned about hard-to-count groups such as immigrants, especially in the Dearborn community of Detroit, which includes one of the most concentrated Arab populations outside the Middle East (Stokes 2020). These groups fit into many groups affected by 2020 census delays: they have experienced heavy impacts from the COVID-19 pandemic, expressed concern about the proposed citizenship question, and been historically underrepresented in census returns (Almulaiki 2021; Alshammari 2020). Community groups such as ACCESS, an Arab American advocacy nonprofit based in Dearborn, strongly pushed census advocacy to ensure groups were counted, but with delayed deadlines and potential undercounts, there is growing concern about representation and COVID-19 support in communities like these (Karoub and Householder 2020).
Based on research from the Center for Urban Research, which produces the Hard to Count Census Map, self-response rates during the non-response follow-up (NRFU) period of the 2020 census improved in underrepresented communities made up of predominantly Black, Hispanic, or Asian groups, in addition to some rural and American Indian communities (Romalewski 2020; CUNY n.d.b). However, response rates in many of these communities still tend to be far lower than in non-Hispanic White communities, with most of these census tracts ending up with response rates lower than in 2010. This information suggests that communities of color needed more door-knocking than non-Hispanic White communities and that the greatest risk of undercounting and data quality concern falls on communities of color.
Again looking to the Center for Urban Research’s report on the response rates in the 2020 census, the final objective of our report is to identify best practices for determining the accuracy and fairness of the 2020 count of the population and the demographic characteristics therein (CUNY n.d.b). The first suggestion from advocates of census integrity is to allow states flexibility to move their constitutionally mandated redistricting deadlines to allow the Census Bureau to provide the most accurate data. Even in the states highlighted here, where voters have enacted specific reforms to provide more transparent and fair mapping processes, proposed changes to deadlines have been met with roadblocks and political disagreement. The ability to delay would allow for districts that best represent the people living in them, while still maintaining periods of public comment that allow for community input. Further, the redistricting processes enacted in states like Michigan and Ohio address key problems with partisan redistricting, even though they are still disrupted by Census Bureau delays. Census data quality is quite difficult to quantify: simple metrics such as the Census Bureau’s touted 99 percent completion rate, or even the aforementioned response rates and NRFU metrics, do not entirely capture the quality of a census count (Tucker 2020). In response to quality issues arising in 2020, multiple advisory groups have proposed improvements to these quality indicators. The American Statistical Association (ASA) convened a 2020 Census Quality Indicators Task Force, which includes 13 census and survey methodology experts, to disseminate a report recommending quality indicators for future census efforts (CQITF 2020). The JASON advisory group, a collection of scientists who advise the US government on science and technology issues, also issued a report assessing the 2020 census data quality processes (JASON 2021). The ASA report outlines multiple recommended metrics that provide improvements to existing checks of US census data quality. These include leveraging extensive self-response data, designing NRFU indicators based on actual NRFU tallies and data collected during the NRFU process, and further implementing post-data collection processing methods. The task force recommends summarizing each of these metrics across geographic areas, in addition to comparing counts from the 2020 census with external estimates. To improve future censuses, the report concludes with further recommendations about the census process: opening the analysis process to qualified external researchers, planning for future censuses in public and leveraging extensive stakeholder input, and conducting additional assessments after collecting more data (e.g., the Post Enumeration Survey). If implemented, these recommendations will help both understanding of the final quality of 2020 census data and improvement of future census processes. The JASON report reached similar conclusions, leading with a recommendation to summarize various assessments of data quality across geographies and demographic groups. It also suggested a comparison of the 2020 census and 2019 census test self-response rates, along with the development of a time series to track changes in enumeration conditions and response rates. Each of these, along with the ASA group’s suggestions, provides an excellent framework to establish the quality of the 2020 census data. The JASON report also suggested that future census work focus on incorporating administrative records and third-party data into census estimates. The Census Bureau used administrative records—data that the federal government already maintains (for example, tax records managed by the Internal Revenue Service)—in new ways during the 2020 census (Mule 2021). These records may significantly lessen the burden of census officials’ future efforts. The Census Bureau also administers more than 130 other surveys and censuses annually, including the American Community Survey (ACS), which provides detailed information about people living in the United States (USCB n.d.). States such as Illinois are using the ACS, among other data sources, to draw legislative maps in lieu of census data (Meisel 2021). There is also opportunity for the Census Bureau to leverage these data sources to measure and improve the quality of future censuses. From the changing deadlines caused by the COVID-19 pandemic, to unqualified appointments to the Census Bureau from the previous administration, to the years-long debate over the citizenship question, many experts are worried that the legacy and the prestige of the census as an institution have been damaged (Bahrampour 2021). These factors, coupled with research suggesting that the Census Bureau’s plans to protect 2020 census respondents’ privacy, may significantly and negatively affect redistricting fairness (Kenny et al. 2021). Taking this all into account, the Census Bureau has much to consider when reflecting on the 2020 census, including better summarizing and communicating data quality, incorporating other data sources, maintaining trust in the Bureau as an institution, and providing usable data while ensuring differential privacy.
Coleman Harris _is a PhD candidate candidate in the Department of Biostatistics at Vanderbilt University. He completed this analysis through a Science Policy Fellowship program with the Center for Science and Democracy at UCS. The Science Policy Fellowship programs is designed to allow current PhD students in diverse fields to gain experience in public policy.
This report relied on discussion and insight from the following experts: John Thompson, Former Director, US Census Bureau; Distinguished Institute Fellow, Biocomplexity Institute and Initiative, University of Virginia Katherine Wallman, Former Chief Statistician of the United States Tom Louis, Former Chief Scientist, US Census Bureau; Professor Emeritus, Biostatistics, Johns Hopkins Bloomberg School of Public Health Mary Jo Hoeksema, Director of Government Affairs, Population Association of America and Association of Population Centers; Co-Director, The Census Project Steve Pierson, Director of Science Policy, American Statistical Association Daniel Elchert, Alfred P. Sloan Foundation Congressional Science Fellow Catherine Turcer, Executive Director, Common Cause Ohio Linda Wurm, Consultant, League of Women Voters of Ohio Jim Kosteva, Voters Not Politicians; Retired Director of Community Relations at University of Michigan