The 2020 Decennial Census: Assessments to
March 24, 2022
Date of Census Quality
Jennifer D. Williams
The 2020 decennial census was innovative in several ways. New technology was developed to
Specialist in American
make updating the Census Bureau’s Master Address File, used for contacting the public, simpler
National Government
and less expensive, and to streamline field operations for collecting data from those who did not

answer the census initially. For the first time, people could answer online as well as on paper
census forms or by telephone. The Census Bureau spent more than $500 million, partly for

extensive advertising, to reach and inform all U.S. residents about the census and encourage them
to complete it.
Despite these efforts, the census encountered critical problems. The Coronavirus Disease 2019 (COVID-19) pandemic
necessitated shutting down, then restarting, census operations. They included the bureau’s special operations to enumerate
hard-to-count people living in group quarters, in shelters, at campgrounds and other transitory locations, and outdoors, plus
people visiting soup kitchens and food vans. Apart from special operations, the bureau’s largest field operation, nonresponse
follow-up (NRFU), called for repeatedly visiting or phoning households that were disinclined to provide their census
responses. In-person NRFU visits, too, had to be postponed.
The Census Bureau reported to the Government Accountability Office a national enumeration rate of 99.98% of housing
units on October 15, 2020. The occupants of 99.0 million housing units (67.0% of the total enumerated) responded by
internet, on paper, or by phone, before NRFU began; the occupants of an additional 48.6 million housing units (32.9%, from
preliminary data) were counted during NRFU.
These numbers told only part of the story. The highest-quality census responses, the bureau noted, came from households that
answered before NRFU or with enumerators’ assistance during NRFU. Preliminary data indicated that 116.1 million
households (about 79%) were in this category, while the remaining 31.04 million households (approximately 21%) did not
complete their questionnaires in the initial response phase or during NRFU. For these households, the bureau resorted to
other means of enumeration—proxy responses, for about 7.4 million occupied households (24.1%) in the NRFU workload;
partial responses, from a yet-to-be-determined number of households; administrative records, for about 8.4 million
households (14.0%) covered by NRFU; and imputation, for an unspecified number of households. A proxy response was one
supplied by a neighbor, building manager, landlord, or someone else presumably knowledgeable about the household’s
composition. Imputation generally involved inferring one household’s characteristics from those of other households nearby.
Group quarters, like college and university housing, proved difficult to count. The bureau had to complete about 1.2 million
NRFU cases by phone instead of in person, a process that did not always go smoothly.
The bureau has conducted a Post-Enumeration Survey (PES), its principal means for assessing census quality, and has begun
releasing the results. Among other important data, the PES yields estimates of differential net undercounts—the extent to
which the 2020 census, like its predecessors, may have undercounted groups such as Blacks and Hispanics relative to non-
Hispanic Whites. Other census quality assessments have been or are being made by the American Statistical Association;
JASON, an independent science advisory group; and the National Academies of Sciences, Engineering, and Medicine’s
Committee on National Statistics.

Congressional Research Service

The 2020 Decennial Census: Assessments to Date of Census Quality

Introduction
More than a decade of planning and testing went into preparations for the 2020 census. Among
the innovations for this enumeration were a first-ever online census response option, use of
governmental administrative records and satellite imagery to eliminate some of the fieldwork
involved in updating census addresses and maps, and greater reliance on technology to streamline
census field operations.1 The Census Bureau conducted extensive advertising, part of a more than
“$500 million public education and outreach campaign,” to publicize the 2020 census and
encourage all U.S. residents to complete it.2 The advertisements were in English as well as 46
other languages that, according to the bureau, reached “over 99% of households more than 300
times.”3 The bureau accepted census responses online, on paper census forms, and by telephone.
Households that needed language assistance to respond online or by phone could receive it in 13
languages. In addition, help was available through guides in 59 languages “in print and video on
2020census.gov.” The bureau mailed households up to seven “invitations” and reminders to
answer the census, including up to two questionnaires.4 Enumerators either were to leave
questionnaires with households lacking home mail delivery or visit them to collect census data in
person. The bureau planned special operations to enumerate hard-to-count people living in group
quarters, in shelters, at campgrounds and other transitory locations, and outdoors, plus people
visiting soup kitchens and food vans. The bureau also was to try repeatedly to visit or phone
households that were disinclined to provide their census responses.5
Some considered these efforts to be inadequate and the census to be seriously flawed.6 The
COVID-19 pandemic, which was beyond what the bureau could have foreseen or encompassed in
its plans, necessitated shutting down, then restarting, census field operations, delaying data
collection7 and causing some confusion about where certain respondents, notably college and
university students, were to be counted.8 The Trump Administration’s attempts to put a

1 For background information on the 2020 census, see CRS Report R44788, The Decennial Census: Issues for 2020, by
Jennifer D. Williams; CRS Report R46237, The 2020 Census: Frequently Asked Questions, by Jennifer D. Williams;
CRS In Focus IF11015, The 2020 Decennial Census: Overview and Issues, by Jennifer D. Williams; and CRS In Focus
IF11486, 2020 Census Fieldwork Delayed by COVID-19, by Jennifer D. Williams.
2 U.S. Census Bureau, “2020 Census Advertising Campaign Expands to Reach All Audiences,” press release CB20-
CN.13, February 18, 2020, at https://www.census.gov/newsroom/press-releases/2020/2020-ad-campaign-expands.html.
3 U.S. Census Bureau, “2020 Census Data Quality,” at https://www.census.gov/programs-surveys/decennial-census/
decade/2020/planning-management/process/data-quality.html#ensuring.
4 Ibid.
5 Ibid.
6 See, as examples, Jonathan Tilove, “Robert Santos Warned That the 2020 Census Would Be Sabotaged. Now He’ll
Oversee the Bureau,” Texas Monthly, November 16, 2021, at https://www.texasmonthly.com/news-politics/robert-
santos-census-bureau/; NPR News, “How many people of color did the 2020 census miss? COVID makes it harder to
tell,” October 9, 2021; Tara Bahrampour, “First batch of data from troubled census set for release,” Washington Post,
April 26, 2021, p. A3; and Tara Bahrampour, “2020 Census may have undercounted Black Americans, new analyses
say,” Washington Post, October 13, 2021, at https://www.washingtonpost.com/dc-md-va/2021/10/13/2020-census-
black-undercount/. A report by a group of nonprofit organizations with interests in the census and civil rights, looking
ahead to 2030, detailed numerous ways in which it believed the 2020 census could have been improved. See New
Venture Fund, Looking to 2030: Findings and Recommendations from Census 2020 Partners and Funders, October
2021, at https://censusproject.files.wordpress.com/2021/11/looking-to-census-2030-full-report-reduced.pdf.
7 See CRS In Focus IF11486, 2020 Census Fieldwork Delayed by COVID-19, by Jennifer D. Williams.
8 D’Vera Cohn, “It’s clear where college students are counted in the 2020 census, but not others who moved due to
COVID-19,” Pew Research Center, April 17, 2020, at https://www.pewresearch.org/fact-tank/2020/04/17/its-clear-
where-college-students-are-counted-in-the-2020-census-but-not-others-who-moved-due-to-covid-19/; and Mike
Schneider, “College towns plan to challenge results of 2020 census,” ABC News, October 18, 2021, at
Congressional Research Service

1

The 2020 Decennial Census: Assessments to Date of Census Quality

citizenship question on the census form, just as it was being readied for printing and posting
online, reportedly raised concerns, including among some of the foreign-born population,
especially noncitizens and people lacking legal status in the United States, and may have further
complicated the already difficult task of persuading them to answer the census.9
Overall delays in census data collection and processing resulted in late releases of the data for
apportioning seats in the House of Representatives and for within-state redistricting. The House
apportionment numbers, which, under Title 13, U.S. Code, Section 141(b), were due no later than
December 31, 2020, were released four months later, on April 26, 2021. They showed an
apportionment population of 331.1 million. This population included residents of the 50 states as
well as members of the U.S. military and federal civilian employees stationed abroad, plus their
dependents living with them. The resident population of the 50 states and DC was reported as
331.4 million.10 Under Title 13, Section 141(c), the tabulations of census data for use in
redrawing state congressional and legislative districts had to be, but were not, delivered by March
31, 2021, to the states that had requested them. The 2020 census redistricting files in a “legacy
format” were released on August 12, 2021, and the same data were released in a somewhat more
usable format on September 16, 2021.11
This report presents selected assessments of 2020 census quality available thus far. Sources are
the Census Bureau itself; the Government Accountability Office (GAO), which typically
publishes reports and makes recommendations regarding the decennial census; the American
Statistical Association (ASA); JASON, an independent science advisory group; and the National
Academies of Sciences, Engineering, and Medicine’s Committee on National Statistics
(CNSTAT), which also generally examines decennial census quality.
Early Assessments by GAO
As the Census Bureau ended 2020 census data collection on October 15, 2020, it reported to
GAO a national enumeration rate of 99.98% of housing units.12 The occupants of 99.0 million
housing units (67.0% of the total enumerated) responded on their own initiative—by internet, by
phone, or on paper census forms—before the nonresponse follow-up field operation (NRFU)
began; the occupants of an additional 48.6 million housing units (32.9% of the total enumerated,
from preliminary data) were counted during NRFU.13 In NRFU, GAO wrote, enumerators were to

https://abcnews.go.com/Lifestyle/wireStory/college-towns-plan-challenge-results-2020-census-80646102.
9Anita Kumar and Caitlin Oprysko, “Trump Abandons Effort to Add Citizenship Question to Census,” Politico, July
11, 2019, at https://www.politico.com/story/2019/07/11/trump-expected-to-take-executive-action-to-add-citizenship-
question-to-census-1405893; Jill Colvin, Mark Sherman, and Zeke Miller, “Trump abandons bid to include citizenship
question on census,” Associated Press, July 11, 2019, at https://apnews.com/article/immigration-donald-trump-ap-top-
news-courts-supreme-courts-18db0fba2743496daeb27a92915bb260; Michael Wines, “2020 Census Won’t Have
Citizenship Question as Trump Administration Drops Effort,” New York Times, July 2, 2019, at
https://www.nytimes.com/2019/07/02/us/trump-census-citizenship-question.html; and Mike Schneider, “Census
analysis finds undercount but not as bad as predicted,” Associated Press, November 2, 2021, at https://apnews.com/
article/coronavirus-pandemic-health-census-2020-think-tanks-congress-5a325059f0a7c7a0d2483b0130551ba9.
10 U.S. Census Bureau, “2020 Census Apportionment Results,” April 26, 2021, at https://www.census.gov/data/tables/
2020/dec/2020-apportionment-data.html.
11 U.S. Census Bureau, “Decennial Census P.L. 94-171 Redistricting Data,” September 16, 2021, at
https://www.census.gov/programs-surveys/decennial-census/about/rdo/summary-files.html#P1.
12 U.S. Government Accountability Office, 2020 Census: The Bureau Concluded Field Work but Uncertainty about
Data Quality, Accuracy, and Protection Remains, GAO-21-206R, December 2020, p. 4.
13 Ibid.
Congressional Research Service

2

The 2020 Decennial Census: Assessments to Date of Census Quality

confirm vacant and nonexistent addresses and make “repeated visits to occupied homes to capture
full information from each household.” If an enumerator could not obtain information from a
household, the enumerator was to leave a “notice of visit” form asking that someone in the
household respond to the census online or by calling a Census Questionnaire Assistance Center.14
The bureau told GAO that by October 15, it had completed 99.93% of its NRFU workload,
covering approximately 64.1 million households. The workload included housing units eventually
determined to be occupied, vacant, or nonexistent; cases in which the bureau had completed
interviews but returned to recheck the information it had collected; and cases in which the bureau
had “high-quality administrative records,” consisting of Social Security, Internal Revenue
Service, or previous census records, that it could use for enumeration after visiting a household at
least once.15
According to the bureau, the highest-quality census responses came from households that
answered before NRFU or with assistance from enumerators during NRFU.16 Preliminary data
indicated that 116.1 million households (about 79%) were in this category, while the remaining
31.04 million households (approximately 21%) did not complete their questionnaires in the initial
response phase or during NRFU.17
For the latter households, the bureau resorted to other means of enumeration.18 Examples were
proxy responses, used to obtain data for about 7.4 million occupied households (24.1%) in the
NRFU workload; partial responses, from a yet-to-be-determined number of households;
administrative records, for about 8.4 million households (14.0%) covered by NRFU; and
imputation, for an unspecified number of households. A proxy response was one supplied by a
person, like a neighbor, building manager, or landlord, presumably knowledgeable about the
household’s composition. Administrative records were mentioned above, and imputation
generally involved inferring one household’s characteristics from those of other households
nearby.
A Subsequent GAO Assessment
A later GAO publication focused on data collection problems and workarounds in the 2020
census.19 Although the Census Bureau reported completing more than 99% of its national
nonresponse follow-up workload by October 15, 2020 (see above), partly through the use of
technology, NRFU cases completed ranged from about 94% to 99% in 10 local areas. In the
Shreveport, LA, area, which had the lowest completion rate (93.5%), a category 4 hurricane in
August 2020 disrupted data collection for 82,863 housing units.20 In some hard-to-count areas,
the bureau “increased the use of travel teams to complete cases,” sending about 26,000

14 U.S. Government Accountability Office, Decennial Census: Bureau Should Assess Significant Data Collection
Challenges as It Undertakes Planning for 2030, GAO-21-365, March 22, 2021, p. 3.
15 U.S. Government Accountability Office, 2020 Census: The Bureau Concluded Field Work but Uncertainty about
Data Quality, Accuracy, and Protection Remains, GAO-21-206R, December 9, 2020, p. 4.
16 Ibid., p. 5.
17 Ibid.
18 Ibid., pp. 5-8.
19 U.S. Government Accountability Office, Decennial Census: Bureau Should Assess Significant Data Collection
Challenges as It Undertakes Planning for 2030, GAO-21-365, March 2021.
20 Ibid., p. 11.
Congressional Research Service

3

The 2020 Decennial Census: Assessments to Date of Census Quality

enumerators to other places during NRFU.21 Occasionally, the bureau offered enumerators
financial incentives for working extra hours “while completing a minimum number of cases, and
removed restrictions on overtime.” In March 2021, noted GAO, the bureau “reported $98.4
million in expenses for enumerator awards, $795,000 of which was for travel awards.”22 In some
areas “struggling to resolve cases or facing challenges to accomplish in-person interviews
because of COVID-19 and natural disasters,” enumerators switched to phone interviews. Bureau
data indicated that “enumerators made more than 10 million calls and completed nearly 1.2
million cases by phone.”23 The bureau’s regional leadership had discretion to authorize the phone
interviews, and bureau officials reported to GAO satisfaction with the process “to make contact
and complete census responses” during NRFU.24 Nevertheless, as one census field supervisor told
GAO, “phone enumerations were difficult to conduct remotely” because of NRFU “rules that
required an in-person proxy rather than over the phone after three contact attempts.”25 GAO
informed the bureau about this concern, and the bureau then “clarified for enumerators how they
could use the handheld device to indicate that they could not complete an in-person proxy, or to
complete the proxy interviews later when they returned to the field.”26
Another difficulty encountered during NRFU was “the inability of supervisors to reassign open
cases in a timely fashion. GAO found that census field supervisors did not have the authority to
reassign cases and had to wait for the field manager to make those reassignments.”27 Bureau
officials informed GAO that the bureau “would consider the reassignment of cases” as it moved
toward 2030 census planning.28
GAO also questioned how adequate certain bureau procedures were for monitoring the quality of
NRFU operations, “such as real-time monitoring of enumerator activities by supervisors and
training assessments.” GAO found that the bureau “did not have proper controls in place,
allowing some enumerators to work without having passed the required training assessment.” The
bureau “agreed that additional controls were necessary.”29
The group quarters enumeration was another matter of GAO concern:
The Bureau planned to count individuals living in group quarters, such as skilled nursing
and correctional facilities, between April 2, 2020, and June 5, 2020, but revised those dates
to July 1, 2020, through September 3, 2020. The pandemic made it difficult to count group
quarters. For example, Bureau staff found it challenging to locate a point of contact at some
group quarters because facilities were closed due to the pandemic. Bureau officials told us
that in December 2020 they decided to re-contact more than 24,000 out of approximately
272,000 group quarter facilities to collect data, and that imputation would be used to count
individuals at the remaining facilities still reporting a zero population count.30

21 Ibid.
22 Ibid.
23 Ibid., p. 12.
24 Ibid., p. 13.
25 Ibid.
26 Ibid.
27 Ibid., GAO Highlights.
28 Ibid.
29 Ibid.
30 Ibid. GAO observed that residents of college and university housing were particularly challenging to count. When in-
person student instruction stopped in March 2020 because of the pandemic, students left their campuses. The bureau
reported receiving census data from 88% of student housing, but colleges and universities “did not always provide the
Congressional Research Service

4

The 2020 Decennial Census: Assessments to Date of Census Quality

A further problem census field supervisors reported to GAO was difficulty accessing large, multi-
unit buildings for NRFU. Enumerators “were often turned away” from these buildings “because
of the pandemic,” but access had been difficult even before it. In such instances, enumerators
were left to collect information from alternative sources like administrative records or proxy
data.31
American Statistical Association Report
GAO wrote in December 2020 that the American Statistical Association and the Census Scientific
Advisory Committee “issued numerous recommendations in the fall of 2020, including for the
Bureau to document what it knows in near real-time about the quality of the population counts
that it provides to the President and to Congress.”32
The ASA recommendations to which GAO referred are from a report by the association’s 2020
Census Quality Indicators Task Force.33 On it were two former OMB chief statisticians, three
former Census Bureau directors, other former bureau officials, and academicians with statistical
and census expertise.
At the time the ASA report was issued, in October 2020, the Census Bureau had “released
information on the percentage of ‘completed enumerations’ by state with a goal of reaching 99
percent completion for each state.” ASA pointed out, however, that the percentage of completed
cases did not “suffice to draw conclusions about data quality.” For example, “the tally of
completed enumerations” covered “households counted through a proxy response from a
neighbor, including cases in which the proxy could provide no information beyond a guess of the
number of individuals living in the household. In fact, meeting enumeration goals for a truncated
deadline” increased “the likelihood of operational shortcuts” possibly affecting “the quality of the
count.”

data” to the bureau. The reasons cited were concerns about students’ privacy, insufficient points of contact for some
fraternities and sororities, and some school administrators’ perception that “completing the census was not a priority.”
Ibid., pp. 24-25.
31 Ibid., p. 13.
A Federal Register notice invited public comment on a proposed 2020 Post-Census Group Quarters Review (2020
PCGQR), which is to provide “a mechanism for tribal, state, and local governmental units in the United States and
Puerto Rico, or their designated representatives, to submit a request that the Census Bureau review their population
counts” for group quarters “by block to correct error(s) affecting the inclusion of group quarters” and their residents in
the 2020 census. The notice specified, “No new decennial information products will be created by the 2020 PCGQR.
The Census Bureau will not revise any 2020 Census information products, such as the population counts delivered to
the President for apportionment or the 2020 Census P.L. 94-171 Redistricting Data Files and Geographic Products.” If,
though, the bureau “determines that a submission identifies existing group quarters” and associated population counts,
“then the Census Bureau will accept the records and update [population] counts for the governmental unit for use in
Census Bureau programs, as appropriate. The Census Bureau will issue certified population count corrections, which
governmental units can use for any purpose requiring their official Census counts. The Census Bureau will incorporate
all 2020 PCGQR revisions into the intercensal population estimates and American Community Survey estimates
starting in 2022 and will post the new counts” on its website. U.S. Department of Commerce, Census Bureau, “Agency
Information Collection Activities; Submission to the Office of Management and Budget (OMB) for Review and
Approval; Comment Request; 2020 Post-Census Group Quarters Review,” 86 Federal Register 64896, November 19,
2021.
32 U.S. Government Accountability Office, 2020 Census: Census Bureau Needs to Ensure Transparency Over Data
Quality, GAO-21-262T, December 3, 2020, pp. 6-7.
33 American Statistical Association, 2020 Census Quality Indicators: A Report from the American Statistical
Association, October 2020, at https://www.amstat.org/asa/files/pdfs/POL-2020CensusQualityIndicators.pdf.
Congressional Research Service

5

The 2020 Decennial Census: Assessments to Date of Census Quality

ASA noted, too, that by October 2020, the bureau had released “extensive response rate data” for
“a number of geographic levels.” The data, added the task force, “need to be analyzed to identify
any lack of uniformity across geographic entities such as states, cities, counties, and rural areas,
and population groups, which would indicate inequity of the counts across the U.S.”34
ASA identified two types of quality indicators for field data collection in the census, notably the
nonresponse follow-up operation. One type of indicator is based on paradata about the NRFU
process: how many attempts an enumerator made “to enumerate a NRFU case,” for instance, or
how much time an enumerator spent on interviews. The other type is based on outcomes, like the
number of households “enumerated by asking a neighbor” how many people were living in the
unit and the number “enumerated using administrative records.”35
ASA suggested that “because of the truncated timeframe” for field operations “and the effects of
the pandemic and multiple natural disasters,” the Census Bureau “make the quality assessment
results available to the public at the census tract level,” the third-lowest level of census
geography. The public thus could ascertain in detail “the extent to which some areas” may have
been “counted more accurately than others and determine the data’s fitness for various uses.”36
The task force called establishing these quality indicators “just the first step” and made five
additional recommendations:
 “The indicators should be readily available and used expeditiously to assess the
quality of the 2020 Census,” preferably before release of the apportionment data
and “most certainly” before the states received their redistricting data.37
 “Qualified external researchers should be granted access to the data to help
conduct the analyses.” Usual Census Bureau practice would be to conduct its
evaluations internally. The unusual 2020 census circumstances, according to the
ASA report, made outside expert assessments advisable to promote transparency,
timeliness, and credibility. “Another important component of this research would
be to allow local area experts such as state demographers to review the early
tabulations and help the Census Bureau determine if unexpected discrepancies
are, in fact, computer processing errors.”38
 “Additional assessments should be conducted when more data become
available.” As the bureau did in several past censuses, it conducted a Post-
Enumeration Survey (PES) that is expected to provide “a wealth of information”
about 2020 census quality. The 2010 census coverage measurement program, for
instance, “allowed for extensive analysis of quality,” covering “duplication[s],
undercounts, and erroneous enumerations.” The task force noted that the PES
results were to be available early in 2022 and possibly could be used to make
corrections in the 2020 data that could affect the distribution of federal funds,
especially to areas estimated to have been undercounted.39
 The bureau typically begins planning for the next census before completing the
current census. The task force proposed that early 2030 census planning draw on

34 Ibid., p. 2.
35 Ibid.
36 Ibid. See below, in contrast, JASON’s comments about the need for aggregated data to protect respondents’ privacy.
37 Ibid., p. 13.
38 Ibid.
39 Ibid.
Congressional Research Service

6

The 2020 Decennial Census: Assessments to Date of Census Quality

the lessons of 2020, “be conducted in public, and include extensive stakeholder
input,” particularly in view of “the concerns that have been expressed” about
2020 census quality. ASA wrote that among the challenges this census faced,
which contributed to doubts about quality, were the COVID-19 pandemic,
restructured census field operations, natural disasters, and litigation.40
 The task force advised that the bureau’s authorizing legislation, Title 13, U.S.
Code, be updated. “First,” the task force wrote, “Title 13 needs to align better
with recent Information Quality Act and OMB guidelines.” The act “requires
agencies to conduct pre-dissemination review of their information products,”
during which “each agency should consider the appropriate level of quality for
each of the products that it disseminates based on the likely use of that
information.” Also, “agencies are required to produce measures of quality that
accompany the release of important data,” like the apportionment and
redistricting data. “Adding these requirements to Title 13 would build
confidence” in the bureau as it continues its work. Second, the task force advised
that Title 13 be “examined to determine if further amendments to protect the
integrity and independence of the Census Bureau and the decennial census are
warranted.”41
JASON Report
As the report by JASON noted, the Census Bureau engaged the independent science advisory
group “to help assess” the bureau’s “data quality processes and procedures, and to elicit
recommendations on areas for improved communication.” The report focused on 2020 census
data quality, “compared with what would have been achieved in better circumstances,” and
especially on “the fitness of the data for the constitutional and statutory uses of the census.”42
The process of readying the resident population counts, the apportionment data, and the
redistricting data for release involved extensive census field operations and post-data collection
processing. “Several intermediate internal data products” preceded the release of the
apportionment counts.43 They were
 the Decennial Response File 1 (DRF1), whose production required the Census
Bureau to “determine the universe of housing units and group quarters,” identify
“unique persons,” standardize responses, and assess the quality of cases for
nonresponse follow-up by incorporating the results of field reinterviews;44

40 Ibid., p. 14.
41 Ibid.
42 JASON, JASON Letter Report JSR-20-2N, Assessment of 2020 Census Data Quality Processes, February 8, 2021, p.
2. The report can be accessed at https://www2.census.gov/programs-surveys/decennial/2020/program-management/
planning-docs/2020-census-data-quality-processes.pdf.
For background information about JASON, see Ann Finkbeiner, “JASON—a secretive group of Cold War science
advisers—is fighting to survive in the 21st century,” ScienceInsider, June 27, 2019, at https://www.science.org/content/
article/jason-secretive-group-cold-war-science-advisers-fighting-survive-21st-century.
43 JASON, JASON Letter Report JSR-20-2N, Assessment of 2020 Census Data Quality Processes, February 8, 2021, p.
4. The report can be accessed at https://www2.census.gov/programs-surveys/decennial/2020/program-management/
planning-docs/2020-census-data-quality-processes.pdf.
44 Ibid., pp. 4-5.
Congressional Research Service

7

The 2020 Decennial Census: Assessments to Date of Census Quality

 the Decennial Response File 2 (DRF2), whose production involved removing
duplicate responses and thus “processing the entire nation at once rather than
state-by-state as in DRF1”;45
 the Census Unedited File (CUF), for which the bureau had to integrate
administrative records and determine “the final population count for each
address,” by imputation if necessary;46 and
 the state-level population counts, whose production involved “merging in the
count of the Federally Affiliated Overseas population for each state,” using
“multiple independent methods” to ensure the accuracy of the apportionment
numbers, and “comparing and reconciling final results.”47
At the time the JASON report was written, the Census Bureau was still “working through its post-
processing of the data” and had not “finalized the delivery dates for the constitutional and
statutory data releases.” As JASON pointed out after being briefed by the bureau on February 2,
2021, “several consequences of the pandemic and other events” increased “the importance of
completing and checking all aspects of the processing milestones.” One possible consequence
cited, for example, was that the pandemic could have exacerbated the problem of duplicate census
responses when college and university students were relocated away from campus and off-
campus housing and back to their “primary residences.”48
The bureau shifted its timeline for data delivery several times, first extending it to allow for
adequate data processing and related quality checks, then compressing the timeline, and finally
returning to the previously extended deadline of April 30, 2021, for delivery of the apportionment
numbers.49 JASON recommended that the Secretary of Commerce return to this deadline, “In
accordance with Sec. 4 of President Biden’s Executive Order (‘Ensuring a Lawful and Accurate
Enumeration and Apportionment Pursuant to the Decennial Census’, January 20, 2021).”50
JASON recommended further that the bureau try to promote confidence in a “potential
differential count of the population” by summarizing its “assessments of data quality across
various geographies and for relevant demographic groups.”51 JASON stressed the importance of
summary data. “One important consideration for any public release of information about data
quality is the risk of disclosing sensitive information.” Although “granular analyses” of detailed
geographic or demographic information would be “useful internally,” any data made public “from
these analyses should be aggregated to avoid disclosure risk.”52
The Census Bureau conducted a nationwide 2019 census test to gauge the effects on response
rates of a possible question about the respondent’s citizenship. JASON suggested comparing the
2019 (pre-COVID-19) response rates from the control group that did not get the citizenship
question with response rates from the 2020 census (during COVID-19) to evaluate the
pandemic’s effects, both on “absolute” response rates and on “preferences between mail-in and

45 Ibid., p. 5.
46 Ibid.
47 Ibid.
48 Ibid.
49 Ibid., pp. 5-6.
50 Ibid., p. 6.
51 Ibid., p. 8.
52 Ibid.
Congressional Research Service

8

The 2020 Decennial Census: Assessments to Date of Census Quality

internet responses.”53 Among the “limiting factors” JASON cited for such a comparison were that
only areas with U.S. Postal Service mail delivery could receive the 2019 test forms; the bureau
conducted nonresponse follow-up in the census but not the test; the bureau required identification
numbers for internet responses to the test, but not the census; the bureau provided less language
support for the test than the census; and the test and the census occurred in different months,
when residence patterns may have differed.54 JASON noted that the Census Bureau already
planned to compare response rates from the 2019 census test and the 2020 census to evaluate “the
efficacy of advertising” on responses. February 2022 was the tentative release date for the study’s
findings.55
JASON advised the bureau to “develop time series that clearly summarize changes in
enumeration conditions and response rates” for its “in-person, online, and mail solicitations.”
According to JASON, time series “would highlight that although enumeration conditions varied,
by adjusting operational strategies,” the bureau “was able to avert deficiencies in enumeration
that otherwise would have appeared.”56
In preparation for the 2020 census, JASON observed, the bureau developed “software and
hardware that made it possible not only to monitor the data collection process, but also to adapt
deployment of resources in response to various issues.”57 As examples, the bureau developed a
Survey Operational Control System and a Field Operational Control System that it used “to
distribute workloads to enumerators.” Enumerators uploaded their data with a Field Data
Collection App, which served to “optimize labor assignments.” These tools “allowed for a much
more integrated approach to the NRFU process.” For instance, when someone on the list for
nonresponse follow-up answered the census before a NRFU visit occurred, “the NRFU list was
immediately updated, thus improving efficiency.” JASON recommended that the Census Bureau
“collect and analyze the data from its control system tools” and summarize the data “to develop
indicators of enumeration progress as a function of time that can be compared” to measures of
progress in past censuses.58
JASON noted that another tool, the Census Review Analysis and Visualization Application
(CRAVA), provided “detailed assessments of data quality at the county level for the responses to
all questions asked” and, “for some responses,” assessments “down to the census block level,” the
lowest level of census geography. The bureau, JASON wrote, should use the CRAVA-based
reports “to show how data quality issues were identified and addressed over time” as the various
data products (DRF1, DRF2, CUF, etc.) were generated.59
JASON also pointed out that in processing census data, the bureau “performs a number of checks
for data completeness and consistency and detects potential inconsistencies” it calls “anomalies.”
These inconsistencies, JASON explained, “are not surprising or indicative of serious problems
but are an expected and normal part of the data quality-control processes.” JASON proposed that
the bureau discontinue using the term “anomaly,” which “may raise inappropriate alarm to non-
experts, and communicate openly about the measures used to check for data consistency, the

53 Ibid., p. 9.
54 Ibid.
55 Ibid.
56 Ibid., p. 11.
57 Ibid., p. 12.
58 Ibid.
59 Ibid., p. 13.
Congressional Research Service

9

The 2020 Decennial Census: Assessments to Date of Census Quality

specific issues uncovered, and steps taken to address them.” If possible, this information should
be compared “with the number and type of issues” identified in the 2010 census.60
JASON acknowledged the bureau’s “significant capacity for communication with many
stakeholders, ranging from primary school students to expert data users,” and mentioned
favorably “the partnerships established with numerous organizations” during the 2020 census,
including efforts to communicate with businesses and state and local governments.61
Nevertheless, JASON found “a shortfall” in the bureau’s communication efforts “regarding its
enumeration processes and changes made to these processes for a range of reasons, including
those resulting in improvements.”62 JASON advised the bureau to “develop tiered messaging,”
from several-minute videos to about an hour-long presentation, that increasingly quantifies “the
myriad efforts undertaken to ensure” an accurate 2020 census count.63
Operational Quality Metrics
In line with ASA’s and JASON’s recommendations, the Census Bureau has made public on its
website certain detailed “operational quality metrics” as 2020 census data have become
available.64 Examples are comparisons to alternative measures of the U.S. population with
Demographic Analysis (DA), discussed below, or population estimates; how the bureau
accounted for nonrespondents, including by imputation or, during NRFU, with administrative
records, proxy responses by neighbors, or responses from household members whom the bureau
contacted; and item nonresponse rates, analyzing nonresponses to particular census questions.
The bureau released sub-state summaries of these metrics but, in keeping with data privacy
concerns expressed by JASON, decided against releasing the metrics for lower geographic levels
like counties or census tracts.65
The 2020 Census Coverage Measurement Program
The Census Bureau has noted its “long history of evaluating population coverage in decennial
censuses,” starting with the 1940 census. “Almost everything we know about the size of the
coverage error, trends in census coverage and differences among subgroups of the population
comes from the Census Bureau’s own evaluation programs.”66 “Coverage error” refers to
population undercounts or overcounts in the census.67 The two principal means of evaluating
coverage, discussed below, are Demographic Analysis and dual-system estimation by means of a
Post-Enumeration Survey.

60 Ibid., p. 14.
61 Ibid., p. 15.
62 Ibid.
63 Ibid., p. 16.
64 See U.S. Census Bureau, “2020 Decennial Census Data Quality, Operational Quality Metrics,” at
https://www.census.gov/programs-surveys/decennial-census/decade/2020/planning-management/process/data-
quality.html#metrics.
65 U.S. Census Bureau, “2020 Census Operational Quality Metrics: Sub-State Summaries,” at https://www.census.gov/
newsroom/blogs/random-samplings/2021/08/2020-census-operational-quality-metrics.html.
66 U.S. Census Bureau, “Coverage Measurement,” at https://www.census.gov/programs-surveys/decennial-census/
about/coverage-measurement.2000.html.
67 U.S. Census Bureau, “Methodology for the 2020 Demographic Analysis Estimates,” at https://www2.census.gov/
programs-surveys/popest/technical-documentation/methodology/2020da_methodology.pdf, December 15, 2020.
Congressional Research Service

10

The 2020 Decennial Census: Assessments to Date of Census Quality

Demographic Analysis
Demographic Analysis produces national-level estimates of the U.S. population that are derived
largely, but not entirely, from administrative records. The federal government acquires these
records, such as on births and deaths, for various reasons unconnected with the decennial census.
DA is designed to be independent of the census.
As the bureau has explained, Demographic Analysis is used to produce “estimates of net coverage
error, which are calculated as the percent difference between the DA estimates and the census
counts.”68 The reference date for the present DA estimates is the official Census Day, April 1,
2020. The DA estimates of the U.S. population on this day, by age, sex, broad racial categories
(Black, alone or in combination with other races, and non-Black, alone or in combination with
other races), and Hispanic and non-Hispanic ethnicity, “were developed using a basic population
accounting approach. The main source of data for the births and deaths is the National Vital
Statistics System maintained by the National Center for Health Statistics. The foreign-born
population for all birth cohorts was estimated primarily using a stock method and data from the
American Community Survey. The estimates of the population born before 1945 (ages 75 and
older) were developed using Medicare records.”69 The bureau “produced three sets of official
estimates,” and “a range of estimates (low, middle, and high)” for each set “to account for
uncertainty in the data, methods, and assumptions used for the 2020 DA.”70
The bureau released its first national tables from Demographic Analysis on December 15, 2020,71
ahead of the April 26, 2021, release of the first 2020 census data: the state apportionment counts72
and resident population counts for the states, DC, and Puerto Rico.73 The bureau noted that “a
report detailing the 2020 Demographic Analysis estimates of net overcounts and undercounts”
was planned for later release.74 The December 15, 2020, data gave the following range of
estimates of U.S. population size on Census Day: low, 330,730,000; middle, 332,601,000; and
high, 335,514,000. The actual census resident population count of 331,449,281 fell between the
low and middle estimates.75
In addition, Demographic Analysis produced national-level low, middle, and high estimates of the
population by age, sex, broad racial categories (Black alone/non-Black alone, Black alone or in
combination/non-Black alone or in combination), and Hispanic/non-Hispanic ethnicity by single
year of age and sex.76 The bureau noted that the estimates could be produced only “in limited race

68 U.S. Census Bureau, “Demographic Analysis (DA),” at https://www.census.gov/programs-surveys/decennial-census/
about/coverage-measurement/da.html.
69 U.S. Census Bureau, “Methodology for the 2020 Demographic Analysis Estimates,” at https://www2.census.gov/
programs-surveys/popest/technical-documentation/methodology/2020da_methodology.pdf, December 15, 2020.
70 Ibid.
71 U.S. Census Bureau, “Census Bureau Releases 2020 Demographic Analysis Estimates,” press release CB20-CN.133,
December 15, 2020.
72 U.S. Census Bureau, “Apportionment Population and Number of Representatives by State: 2020 Census,” at
https://www2.census.gov/programs-surveys/decennial/2020/data/apportionment/apportionment-2020-table01.pdf.
73 U.S. Census Bureau, “Resident Population for the 50 States, the District of Columbia, and Puerto Rico: 2020
Census,” at https://www2.census.gov/programs-surveys/decennial/2020/data/apportionment/apportionment-2020-
table02.pdf.
74 U.S. Census Bureau, “Census Bureau Releases 2020 Demographic Analysis Estimates,” press release CB20-CN.133,
December 15, 2020.
75 Ibid.
76 Ibid.
Congressional Research Service

11

The 2020 Decennial Census: Assessments to Date of Census Quality

detail because they relied on historical records and measures of race that have changed over
time.” Further, DA estimates of the Hispanic population were not produced for all ages because
the option for reporting Hispanic ethnicity “was not widely available on birth and death records
until the 1990s.”77
 The low, middle, and high percentages of the U.S. population estimated to be
Black alone were, respectively, 13.5%, 13.7%, and 13.9%.78
 The percentages of the population estimated to be Black alone or in combination
with other races were 14.9%, 15.1%, and 15.4%.79
 The percentages of the population under age 30 estimated to be Hispanic were
23.0%, 24.6%, and 26.0%.80
The Census Bureau has cautioned against prematurely comparing the DA estimates of race and
ethnicity with data from the 2020 census or population estimates: “Many people want to evaluate
how well specific race groups were counted in the 2020 Census by comparing the race data from
the 2020 Census to the DA results or the population estimates. We caution against making these
comparisons right now because the available benchmarks do not have the same race categories as
the census.”81
Post-Enumeration Survey
Like Demographic Analysis, the Post-Enumeration Survey is designed to be independent of the
census and to indicate the extent of census accuracy. Unlike DA, though, the PES is intended to
be a nationally representative sample survey whose results are matched with census results to
estimate census errors both nationally and subnationally. More specifically, the bureau has
explained, the 2020 census PES uses “dual-system estimation,” in which the two systems are the
PES and the census. The PES “independently interviews people, asks where they lived on April 1,
2020, and then matches that information to the census results.” The PES “takes more than two
years and involves enumerating housing units and people from scratch in about 10,000 blocks
across the country.” By matching “these housing units and people to the list of addresses and
people in the census,” the bureau can determine who was counted only in the census, only in the
PES, and in both the census and the PES.82 The PES should enable the Census Bureau to provide
estimates of “net coverage error” as well as “components of coverage.”
“Net coverage error,” in the bureau’s terminology, refers to “the difference between the census
count and the PES estimate of the actual number of people in the U.S.” Negative coverage error
would mean that the census count was “too low and the census missed some people,” so had an
undercount. Positive coverage error, or an overcount, would mean that the census count was “too

77 Ibid.
78 Ibid.
79 Ibid.
80 Ibid.
81 Eric Jensen and Sandra Johnson, “Using Demographic Benchmarks to Help Evaluate 2020 Census Results,” U.S.
Census Bureau, November 9, 2021, at https://www.census.gov/newsroom/blogs/random-samplings/2021/11/
demographic-benchmarks-2020-census.html. See especially, at this site, mention of the 2020 Modified Race File and
the Demographic and Housing Characteristics File.
82 Timothy Kennel, “The Post-Enumeration Survey: Measuring Coverage Error,” U.S. Census Bureau, December 16,
2021, at https://www.census.gov/newsroom/blogs/random-samplings/2021/12/post-enumeration-measuring-coverage-
error.html.
Congressional Research Service

12

The 2020 Decennial Census: Assessments to Date of Census Quality

high, indicating some people” may have been “counted more than once.”83 The bureau,
additionally, checks the census numbers and the PES estimates for various groups according to
their demographic characteristics, namely: groups by age, sex, race (Whites, with a breakout for
non-Hispanic Whites; Blacks; Asians; American Indians or Alaska Natives, broken out by
American Indians or Alaska Natives living on reservations, living on American Indian Areas off
reservations, and living in the rest of the United States; Native Hawaiians or other Pacific
Islanders; and people who identify as some other race), and Hispanic ethnicity. The bureau then
compares “net coverage error rates” across demographic groups to determine differences in how
the census counted them. “When a group has a larger or smaller net undercount than the country
as a whole,” the bureau calls it a “differential net undercount.”84
“Components of coverage” that the bureau was to identify during PES fieldwork consist of
correct enumerations, erroneous enumerations, and whole-person imputations. “Correct
enumerations” in the 2020 census refer to U.S. residents counted as of April 1, 2020, where they
should have been counted, in the United States at their usual residences. “Erroneous
enumerations” include “duplicate records of people who were correctly counted in the census as
well as people who were counted but should not have been. For example, they may have died
before April 1, 2020,” or been just visiting the United States. “Whole-person imputations” refer to
“a statistical technique” used to fill in data missing from census records. The bureau is to count
the number of census records requiring whole-person imputation and use the PES to estimate
correct and erroneous enumerations.85
The Census Bureau has used some type of survey to estimate census coverage and accuracy since
1950.86 The evaluation of the 1980 census was designated the Post-Enumeration Program. In the
1990 census, the Post-Enumeration Survey was known by that term. In 2000, it was called
Accuracy and Coverage Evaluation and in 2010, Census Coverage Measurement.87
Fieldwork for the 2020 census Post-Enumeration Survey, again given that name, began in January
2020, when the bureau started developing an address list independent of the one used for the
actual census.88 The bureau announced on February 11, 2021,89 that it was hiring temporary
workers for the in-person follow-up phase of the PES, to have begun in June. During this phase,
the bureau was to revisit previously interviewed housing units to confirm its records. The first
results of the PES were tentatively scheduled for release at the end of December 2021 and more
findings by the end of March 2022.90 According to an October 9, 2021, NPR news report,
however, in-person PES interviews were interrupted by the pandemic, so were to take place from
November 2021 to February 2022.91 In a November 16, 2021, press release, the bureau reported
that follow-up would begin on November 29, 2021, and continue through March 2022, with

83 Ibid.
84 Ibid.
85 Ibid.
86 Ibid.
87 U.S. Census Bureau, “Post-Enumeration Surveys,” at https://www.census.gov/programs-surveys/decennial-census/
about/coverage-measurement/pes.html.
88 U.S. Census Bureau, “Census Bureau Hiring for Independent Survey That Measures Quality of 2020 Census,” press
release CB21-CN.12, February 11, 2021.
89 Ibid.
90 Ibid.
91 NPR News, “How many people of color did the 2020 census miss? COVID makes it harder to tell,” October 9, 2021.
Congressional Research Service

13

The 2020 Decennial Census: Assessments to Date of Census Quality

release of the first PES results planned for “the first quarter of 2022” and additional results
expected “in the summer of 2022.”92
On March 10, 2022, the bureau released its national-level PES estimates,93 with this overall
assessment:
“Today’s results show statistical evidence that the quality of the 2020 Census total
population count is consistent with that of recent censuses. This is notable, given the
unprecedented challenges of 2020,” said Director Robert L. Santos. “But the results also
include some limitations—the 2020 Census undercounted many of the same population
groups we have historically undercounted, and it overcounted others.”94
The census undercounted Blacks, American Indians or Alaska Natives living on reservations,
Hispanics, and people who reported being of some other race.
The census overcounted non-Hispanic Whites and Asians. Native Hawaiians or other Pacific
Islanders were neither overcounted nor undercounted, the estimates indicate.
Among age groups, the 2020 census undercounted children up to age 17, especially children up to
age 4. “Young children are persistently undercounted in the decennial census,” the bureau noted.95
More specific PES results by race and Hispanic ethnicity follow below.96
 Blacks, alone or in combination with other racial groups, “had a statistically
significant undercount of 3.30%. This is not statistically different from the 2.06%
undercount in 2010.”
 Hispanics “had a statistically significant undercount rate of 4.99%. This is
statistically different from a 1.54% undercount in 2010.”
 American Indians or Alaska Natives, alone or in combination with other groups,
who lived on reservations, had “a statistically significant undercount rate of
5.64%,” not “statistically different from a 4.88% undercount in 2010.”
 “The non-Hispanic White alone population had a statistically significant
overcount rate of 1.64%. This is statistically different from an overcount of
0.83% in 2010.”
 Asians, alone or in combination with other groups, “had an overcount rate of
2.62%. This is statistically different from 0.00% in 2010.”
 Native Hawaiians or other Pacific Islanders, alone or in combination with other
groups, “had an estimated overcount rate of 1.28%. This rate is not different from
an estimated 1.02% overcount rate in 2010.” Neither rate is “statistically different
from zero.”
The PES estimates include components of coverage (mentioned above), namely “the number of
correct census records, erroneous enumerations, whole-person imputations, and omissions.”97 The
DA results do not include coverage estimates by component.

92 U.S. Census Bureau, “Census Bureau’s Final Post-Enumeration Survey Field Operation Set to Begin,” press release
CB21-CN.76, November 16, 2021.
93 U.S. Census Bureau, “Census Bureau Releases Estimates of Undercount and Overcount in the 2020 Census,” press
release CB22-CN.02.
94 Ibid.
95 Ibid.
96 Ibid.
97 Ibid.
Congressional Research Service

14

The 2020 Decennial Census: Assessments to Date of Census Quality

According to the PES, of the 323.2 million people living in housing units on April 1, 2020, an
estimated
 “94.4% were counted correctly”;
 “2.2% were counted erroneously (1.6% were duplicates and 0.6% were erroneous
for other reasons)”;
 “2.8% provided only a census count and had their demographic characteristics
imputed, or statistically inserted”; and
 “0.6% needed more extensive imputation after all census follow-up efforts were
attempted.”
In addition, the PES estimates indicate 18.8 million 2020 census omissions. “Omissions,”
according to the Census Bureau, “are people who were not correctly counted in the census. Some
of these people were missed by the census. However, omissions may have been accounted for in
the census counts as whole-person imputations.”98
PES and DA Estimates Compared
Limited comparisons of the PES and Demographic Analysis findings accompanied the Census
Bureau’s March 10 release of the PES estimates.99
The two sets of estimates produced results by age and sex. For ages 0-17, the PES shows a
statistically significant undercount of 0.84%, and the middle DA series shows a 2.1% undercount.
At ages 0-4, the PES indicates a statistically significant undercount of 2.79%, and DA shows a
middle series undercount of 5.4%.100
The principal difference between the PES and DA results is at ages 18-29, where the PES shows
statistically significant undercounts of 2.25% for males and 0.98% for females. DA, in contrast,
shows overcounts for males and females, except males in the high series. The bureau pointed out
that the DA estimates refer to the total population living in the United States on April 1, 2020,
while the PES estimates are based on the household population living in the United States,
excluding Remote Alaska areas. Further, a considerable number of people in this age category
were living in college dormitories, which were not part of the PES sample. Census Bureau
researchers, the bureau stated, “acknowledge this notable difference” between the PES and DA
estimates and are trying to account for it.101
The PES and DA estimates of net coverage error for adults ages 30-49 show similar patterns. The
PES indicates a statistically significant undercount of males in this age range, and DA also shows
an undercount.102
The PES shows statistically significant overcounts of 0.55% for males and 2.63% for females
ages 50 and older. DA, too, indicates overcounts of males and females ages 50 and older, except
males in the high series.103

98 Ibid.
99 Ibid.
100 Ibid.
101 Ibid.
102 Ibid.
103 Ibid.
Congressional Research Service

15

The 2020 Decennial Census: Assessments to Date of Census Quality

Review by the Committee on National Statistics
The National Academies of Sciences, Engineering, and Medicine’s Committee on National
Statistics has formed a panel “to review and evaluate the quality of the 2020 census.” The review
continues CNSTAT’s long practice of independently assessing census data quality as well as the
operations that produced the data. CNSTAT noted that this review is particularly needed in view
of some “major changes to decades-old processes” and the “unprecedented” challenges to the
census associated with COVID-19.104
The panel is examining the Census Bureau’s information about the census data it collected,
including indicators of data quality obtained during census operations; information from
Demographic Analysis and the PES; analyses of administrative records; and relevant evaluations
from the 2010 and 2000 censuses.105
The panel is to produce an interim report and, by April 2023, a final report with its conclusions
and recommendations for the 2030 census.106

Author Information

Jennifer D. Williams

Specialist in American National Government

Disclaimer
This document was prepared by the Congressional Research Service (CRS). CRS serves as nonpartisan
shared staff to congressional committees and Members of Congress. It operates solely at the behest of and
under the direction of Congress. Information in a CRS Report should not be relied upon for purposes other
than public understanding of information that has been provided by CRS to Members of Congress in
connection with CRS’s institutional role. CRS Reports, as a work of the United States Government, are not
subject to copyright protection in the United States. Any CRS Report may be reproduced and distributed in
its entirety without permission from CRS. However, as a CRS Report may include copyrighted images or
material from a third party, you may need to obtain the permission of the copyright holder if you wish to
copy or otherwise use copyrighted material.

104 National Academies of Sciences, Engineering, and Medicine, Committee on National Statistics, Panel to Evaluate
the Quality of the 2020 Census, at https://www.nationalacademies.org/our-work/panel-to-evaluate-the-quality-of-the-
2020-census.
105 Ibid.
106 Ibid.
Congressional Research Service
R47046 · VERSION 3 · UPDATED
16