Comments on Proposed Information Collection on 2020 Census, Docket # USBC-2018-0005

View a PDF of these comments here.

Dear Ms. Jessup:

On behalf of The Leadership Conference on Civil and Human Rights, a coalition charged by its diverse membership of more than 200 national organizations to promote and protect the civil and human rights of all persons in the United States, and its Census Task Force co-chairs, Asian Americans Advancing Justice | AAJC and NALEO Educational Fund, we appreciate this opportunity to provide comments in response to the Federal Register notice (the “Notice”) seeking comments on significant aspects of the 2020 Census operational plan.

The Leadership Conference provides a powerful unified voice for the many constituencies of the coalition: persons of color, women, children, individuals with disabilities, gays and lesbians, older Americans, labor unions, major religious groups, civil libertarians, and human rights organizations. Our coalition views an accurate and fair census, and the collection of useful, objective data about our nation’s people, housing, economy, and communities generally, to be among the most important civil rights issues of our day. We have a long record of first-hand experience working in support of previous censuses. For the 2010 Census, we undertook the most comprehensive and extensive effort by a stakeholder organization to promote participation in historically hard-to-count communities and to mobilize local advocates in support of the census by highlighting the civil rights and social justice implications and community benefits of an accurate count.

At the outset, we want to amplify the constitutional requirement for a census as the foundation of a democratic system of governance based on equality of representation. Accurate census data also are essential for the prudent, fair allocation of public resources at all levels of government; development of policies and programs that meet the needs of all communities effectively and efficiently; and as a guide for private sector investment that spurs economic growth and creates jobs.

The Census Bureau has taken many steps over the decades to improve the accuracy of the enumeration, and we have appreciated the opportunity to be partners in advancing that goal in recent decades. We also applaud innovation in all federal data collection activities, which we believe encourages creativity and forward-thinking, as well as wise expenditure of taxpayer dollars, and takes advantage of scientific advancements in survey measurement and ever-evolving technologies.  Nevertheless, the census continues to count some communities and population groups more accurately than others, resulting in differential undercounts for people of color, renters (a proxy for lower income households), young children, American Indians living on reservations, and many rural and remote households, compared to non-Hispanic Whites, homeowners, and some older Americans.[1] Therefore, we must evaluate the proposed 2020 Census design, methodology, and operations in the context of an overarching goal to eliminate these persistent differential undercounts. All elements of the 2020 Census plan must increase the likelihood of counting all population groups and all communities equally well and remove barriers to full participation to the greatest extent possible. Cost effectiveness and innovation in census taking are important, worthwhile goals, but they can never override or undermine the Census Bureau’s ability to produce the fair and accurate enumeration of our nation’s population that the Constitution requires.

Issues for Discussion

Along with more than 140 other organizations, we submitted separate comments on August 1 in response to this Notice, urging the Census Bureau in the strongest possible terms to remove the untested, unnecessary citizenship question from the 2020 Census form. Today’s comments address a broader set of design and operational issues and provide a civil rights perspective on selected elements of the census plan set forth in the Notice. We comment on those issues, and offer recommendations when appropriate, in the order in which they are set forth in the Notice.

Reengineering Address Canvassing

We agree that a “complete and accurate address list” is fundamental to a successful, inclusive census. We also recognize that the address canvassing operation has been costly and time consuming in previous decades, making improvements to the process for ensuring a comprehensive Master Address File (MAF) and TIGER mapping database a necessary goal.

1. Type of Enumeration Area (TEA) designations

Census stakeholder organizations and local governments involved in “Get Out the Count” campaigns must be able to convey accurate information about the basic census process — when and how households will receive materials related to census participation — to the communities they serve. To our knowledge, the Census Bureau has not released information on Type of Enumeration Area designations for all communities in the country. In 2010, confusion in some areas — including the Rio Grande Valley — about the method for delivering questionnaires shook confidence in the census process; national and grassroots advocates had to help the Census Bureau restore effective lines of communication in these hard-to-count communities. We cannot afford a repeat of that sort of communications breakdown.

Recommendations:

• We urge the bureau to make TEA designations for all communities available publicly as quickly as possible, to help its governmental and organizational partners develop their educational materials and plan their outreach activities in a timely way.

• The Census Bureau should mail a letter or postcard to post office boxes in areas designated as Update/Leave or Update/Enumerate TEAs before the start of the national self-response period, to alert residents about the method of delivery of their census materials. In addition, the bureau should make educational materials (such as fliers and posters) available at Post Offices in these TEAs, informing residents about census methods and timing applicable to them.

1. Address Canvassing

Undoubtedly, both automation and the proliferation of government and commercial databases that identify a significant portion of the nation’s housing units will help the Census Bureau use its limited fiscal resources more prudently. In areas dominated by single family homes and stable multi-unit buildings, where growth in the housing unit stock is limited or easily documented through high quality administrative data and aerial imagery, and also updated through Geographic Support programs such as the Local Update of Census Addresses (LUCA), the bureau can review and verify the MAF and TIGER maps in more cost effective ways than in past censuses.

However, we remain concerned that long term and ongoing fiscal pressures have led the bureau to limit and even further reduce the scope of In-Field Address Canvassing, thereby jeopardizing the bureau’s ability to locate or verify the existence of nontraditional living quarters and hidden housing units, including subdivided single family structures, converted or shared commercial spaces, garages and sheds, and campers located on personal property. We note, for example, that the bureau’s Fiscal Year 2019 budget request projects a need for 76,000 address listers, a significant reduction from the original estimated address canvassing workforce of 90,000 — even as the percent of housing units assigned to the In-Field canvassing operation increased by five percent (from 25 to 30 percent), and the scope of In-Field canvassing in the three 2018 End-to-End Census Test sites actually ranged from 37 to 76 percent of the relevant portions of the MAF.[2]

Harder-to-find living quarters are home to individuals and families who already are considered hard-to-count. Failure to dedicate adequate staff and resources to verify and improve the accuracy of the MAF in the field could well contribute to a higher undercount and larger differential undercount of population groups historically at risk of being overlooked, by virtue of “whole household” omissions.

Recommendations:

• The Census Bureau should devote more resources to conducting In-Field Address Canvassing and expand the scope of on-the-ground canvassing to increase the likelihood of capturing nontraditional, hidden, and converted housing units in the MAF before the census starts.
• The bureau should consult with community-based organizations familiar with neighborhoods and rural areas where hard-to-find housing is prevalent, both as a source of a knowledgeable hiring pool for the address canvassing operation and for guidance on the types of living quarters address listers might overlook. Equally important, residents of many neighborhoods and areas where address canvassing will take place are likely to view address listing staff with suspicion and even fear, equating any government personnel who are walking around their communities with federal and local law enforcement authorities. Stakeholder organizations can serve as trusted messengers in communities where address canvassing is taking place, reassuring residents about the purpose and safety of the work taking place.

Optimizing Self-Response

1. Internet Self-Response

As we recognized at the outset, it is an important and worthwhile goal for the Census Bureau to explore uses of new and emerging technologies to facilitate response to the census and improve cost- efficiencies as the bureau seeks to contain the cost of the enumeration at the direction of Congress. We note that the Census Bureau successfully introduced an internet response option in the American Community Survey (ACS) in 2013.

The internet clearly offers a less expensive way for people to respond to the census; a higher internet response rate could, absent unforeseen failures in other census operations, make additional resources available to reach hard-to-count population groups and communities and facilitate their direct participation. The internet also offers the opportunity for people who did not receive, retain, or have access to the unique code associated with their home address to submit their census responses, to a greater degree and in a less costly way than previous “Be Counted” campaigns allowed.

A high-tech census, however, introduces cyber-security risks, both known and unknown. As we have seen with hacking incidents affecting organizations and institutions from government agencies, to hospitals, to corporations, and our electoral system, the threat of a cyber-breach involving domestic or foreign actors is real. Of equal concern is the level of public confidence in the security of the Census Bureau’s IT systems and databases, which likely will affect people’s willingness to respond on-line. Real or perceived, cyber-security threats introduce uncertainty into the entire census process and could invalidate cost assumptions and stretch resource availability in real time. The Census Bureau must be thoroughly prepared to address all possibilities related to a breach of its IT systems — or diminished public confidence stemming from a highly visible, if unrelated, data breach affecting another U.S. institution — quickly, decisively, and seamlessly, to avoid irreparable disruption to the enumeration.

We also are deeply concerned about the likelihood that an internet response option makes households vulnerable to becoming victims of phishing scams through email or texting applications, as well as of fake websites purporting to be official portals for on-line response. Likewise, a well-timed misinformation campaign by malicious actors can further cause individuals to decide not to respond or respond inappropriately.  The Census Bureau, working closely with governmental and external technical experts and its communications campaign contractors, must do everything possible to protect the public against these scams.

A third and no less significant concern related to a census plan that emphasizes internet response over other means of participation are the implications of the digital divide for ensuring equitable opportunities for households without reliable broadband and Internet access to be counted. It simply is not enough to say that the 2020 Census will treat all communities fairly and equitably because people will have multiple ways to provide their responses, if those methods are not equally accessible and robust. For example, some respondents may only be able or willing to participate in the census using a traditional paper questionnaire, yet paper forms will be available only in English and Spanish, compared to 13 languages for on-line and telephone response modes.

In 2017, The Leadership Conference Education Fund and the Georgetown Census on Poverty and Inequality addressed a full range of challenges and concerns related to a high-tech census in a report entitled, “Counting Everyone in the Digital Age.”[3] We commend the full report to you, which includes detailed information on marginalized communities that lack reliable internet access, among other important research findings. We are pleased to highlight the following recommendations from the report, updated to reflect refinements in the 2020 Census plan. Additional recommendations related to automation of census operations are included in our comments on Reengineering Field Operations, below.

• The Census Bureau should take all possible measures to bridge the digital divide with the new internet response option. For instance, the bureau should test software for the decennial census to ensure it operates smoothly and safely on older operating systems and smartphones that may not handle new software and new websites well. This is especially important since people of color, younger adults, low-income adults, and those without a high school diploma are most likely to use older computers or a mobile device as a primary source of internet access. The bureau also must address the significant digital divide affecting many rural and remote communities, including American Indian reservations and Alaska Native villages, and ensure that there are adequate resources to conduct a comprehensive enumeration in areas where reliable broadband is problematic.
• Both the Census Bureau and the Department of Homeland Security (DHS) should guarantee the sanctity of personal data submitted as part of the 2020 Census. DHS should officially pledge to strip all personally identifiable information from any cybersecurity threat indicator it shares outside the department. With respect to Einstein 3A intercepted traffic, there is a significant threat that, absent a complete scrub of personally identifiable information from any cyber threat indicator, personal information could be used, and access abused, for purposes that will chill participation in the census.

• The Census Bureau should partner with schools, libraries, community centers, and other public computing centers that can set aside computers to facilitate on-line responses to the census for households without reliable broadband or internet access. Doing so during the self-response period would give more people an opportunity to be counted on systems that are updated, secure, and using software that optimizes response time. To this end, the Census Bureau must avoid any policy that limits the number of self-responses its IT system will accept from a single IP address or electronic device, as such limits would undermine efforts the bureau’s 2020 Census staff themselves — as well as state and local governments — are promoting to facilitate internet response in group settings whenever helpful.

• The Census Bureau should work with local partners, including civil rights advocates, to emphasize the confidentiality of responses and help respondents sift through perceived cybersecurity and privacy threats to determine legitimate strategies and outstanding concerns. The communications campaign, as well as trusted messengers who will be a conduit to hard-to-count communities, should explain clearly how to identify official on-line decennial census materials, forms, and portals, and how to avoid phishing attempts and fake websites, especially since the Census Bureau is using targeted digital advertising.

Utilizing Administrative Records and Third-Party Data

In a May 2017 report, the Urban Institute, with support from The Leadership Conference on Civil and Human Rights and the Georgetown Center on Poverty and Inequality, considered the numerous policy and operational issues associated with the proposed uses of government and commercial data in the 2020 Census. “Administrative Records in the 2020 US Census: Civil Rights Considerations and Opportunities”[4] documents challenges and concerns about the use of these data to: (1) build and update the Master Address File; (2) determine the occupancy status of unresponsive housing units and remove vacant and nonexistent units from the NRFU workload; (3) optimize the work of field staff during the NRFU operation; and (4) enumerate some households that do not self-respond. The report looks at the costs, benefits, and risks of relying on administrative data, to varying degrees, for each of these major components of the census. Again, we commend the full report to Census Bureau staff working on each of these important 2020 Census operations. We also offer the additional observations and recommendations below.

1. In-Office Address Canvassing

We previously offered general support for the use of administrative data to assist with verifying and updating the residential address file. At the same time, we urge caution with respect to the limits of these datasets, which are far less likely to include accurate information on nontraditional and hidden housing units that, more often than not, are home to individuals and families already considered hard-to-count based on a range of demographic characteristics identified in previous censuses and 2020 Census pretests. Over reliance on government and commercial data to build an accurate address list undoubtedly will result in failure to count a disproportionate number of households in historically undercounted communities.

1. Identifying vacant and non-existent housing units

We agree, in theory, that administrative (and possibly third-party) data can help streamline census field work by identifying vacant and nonexistent housing units before the start of the Nonresponse Follow-up (NRFU) operation, thereby saving resources and staff time that can be redirected to successful enumeration attempts at occupied housing units. However, we raise a note of significant caution with respect to the accuracy of vacancy status information in the datasets the Census Bureau proposes to use. Evaluation of census field test results showed that roughly 25 percent of housing units identified as vacant based on U.S. Postal Service data were, in fact, occupied. This potential error level is unacceptable, and the bureau must conduct a similar evaluation of vacancy status determinations in the 2018 End-to-End Census Test and release the results publicly before finalizing plans to remove vacant units from the NRFU universe using administrative records, without an in-person visit to confirm that status. Further, we are concerned that the Postal Service’s inability to deliver census materials to an address on the MAF may not accurately reflect vacancy status of a residential structure. Alternative explanations for UAA mail are residents who receive their mail only at a post office box, instead of at their home address, and living quarters with unclear, unique addresses not recognized by the USPS. The Census Bureau must also ensure that its procedures for removing supposedly vacant housing units from the NRFU operation do not eliminate all opportunity to identify structures (often abandoned, especially in low income neighborhoods) where people experiencing homelessness live, perhaps for long periods of time.[5]

Recommendation:

• The Census Bureau should develop an additional procedure to prevent overlooking people who reside in housing units that are officially designated as vacant or abandoned and to enumerate these residents. Such a procedure could involve consultation with knowledgeable community-based partner organizations, who can identify specific census blocks or neighborhoods where such living arrangements are likely to occur.

1. Enumeration of unresponsive occupied housing units

We are extremely concerned about the proposed use of administrative records to enumerate occupied households that do not self-respond, and we have not yet seen information to confirm that the benefits associated with this proposed methodology outweigh the risks in terms of less accurate and lower quality data. By definition, households in the NRFU universe are hard-to-count, and therefore more likely to be missed, because they failed to self-respond. That means that any weaknesses in the methodology will disproportionately affect certain population groups, which could undermine the fairness of the census results.

The Notice appears to narrow the range of datasets the Census Bureau will use to determine if records of sufficient quality are available for this purpose, since previous research and testing included administrative data from several state-administered programs. This modified plan could remove the possibility that the Census Bureau would not have consistent access to administrative records for all states and the District of Columbia, as well as the significant challenge of ensuring consistent quality, definitions, and timeliness of the data from state to state. Nevertheless, we do not believe that federal administrative records include sufficient information about the race and ethnicity of household members, especially at the level of granularity (subgroup or origin, for example) the bureau seeks to achieve in this census. We also are skeptical that these records can establish accurate information about household relationship, which is an important variable for policymakers and researchers seeking to understand the needs of families and nonfamily households or to identify nonfamily household members who do not have a usual home elsewhere, even if they do not intend to stay at the home permanently. Furthermore, administrative records generally do not cover undocumented residents. Research also has shown that many government datasets do not cover young children accurately, a significant concern given the historical disproportionately high undercount of children ages 0 – 4.

Recommendations:

• The Census Bureau should make more information available that demonstrates the quality thresholds it has established to determine if administrative records will be used to enumerate occupied unresponsive households after only one in-person visit and a follow-up postcard. Those thresholds must include race and Hispanic origin data consistent with the categories offered on the 2020 Census questionnaire. The bureau also must publicly address how it will overcome the deficiencies in administrative records noted above, with respect to transitory nonfamily household members, young children, and undocumented residents, among other hard-to-count individuals.
• Administrative records should be used as a source of information for unresponsive households only after a minimum of three unsuccessful in-person visits and in conjunction with data gleaned from proxy attempts.

Reengineering Field Operations

4. Automation

We commend the Census Bureau for pursuing opportunities to automate the work of census enumerators, given a similar unsuccessful effort for the 2010 Census. When feasible and sound, the bureau should take advantage of new technologies to streamline and improve the cost-effectiveness of census operations. However, in field tests, the bureau has encountered various difficulties with the mobile devices that field staff will use to collect information during the Address Canvassing, NRFU, and quality control operations, including lack of consistent connectivity in rural areas, overheating of devices in areas experiencing very high temperatures, and software malfunctions. The bureau must have contingency and back-up plans to ensure that address listers, enumerators, and other field staff can continue to carry out their duties in a timely way if mobile devices do not function as expected.

We also are concerned that the application process for census positions will largely be limited to the internet. Data clearly show that historically hard-to-count population groups are less likely to have reliable broadband (especially in rural areas and on American Indian reservations) or access to the internet. Older Americans, who often represent an important source of qualified, available census workers, also are far less likely to use the internet. Further, we have heard troubling reports from individuals in proximity to the 2018 End-to-End Census Test sites that the on-line application process is burdensome and time-consuming, initially because of the requirement to create an account through the <usa.jobs> portal.

Recommendation:

• The Census Bureau should ensure adequate opportunities for people to apply for census positions using traditional paper applications, and also facilitate convenient on-line applications for individuals who do not have access to the internet.

1. Field offices

We are greatly concerned that automation of field operations has resulted in an insufficiently robust field infrastructure plan. Fiscal constraints earlier in the decade reduced the number of permanent regional census offices from twelve to six, which in turn, we believe, has driven the decision to open only six regional census centers (RCC) — half the number that were established during the 2010 Census. This reduction means that RCC directors and management staff are responsible for significantly larger geographic areas with a greater diversity of communities and populations, thus making it more difficult for senior staff to establish strong relationships with all jurisdictions and key leaders within their territories. It therefore will be harder to implement effective rapid response activities in a timely way when unanticipated challenges arise.

Similarly, the 2020 Census plan includes only 248 Area Census Offices (ACOs), roughly half the number of local census offices in the 2010 Census. The bureau determined the number and location of ACOs using a formula that considered factors such as projected workload for NRFU and other field operations (e.g. Group Quarters, Service-based Enumeration, Update/Enumerate). Yet when the Secretary of Commerce revised the census lifecycle cost estimate in October 2017, based in part on a lower self-response rate and larger NRFU universe, the bureau did not revise its field infrastructure plan to meet the increased workload. With the self-response rate likely to be even lower than current projections due to the proposed citizenship question, we believe the bureau must plan for more local offices, to ensure that its managers can address difficulties during field operations with a sufficient level of knowledge about the communities covered by their offices. The significantly smaller field infrastructure removes an important element of successful census-taking in hard-to-count communities: knowledge of on-the-ground conditions, challenges, cultural norms and potential cultural barriers, and community and elected leaders. Automation, we believe, must support and enhance a sound census design, not serve as a rationale for cutting costs in ways that undermine the inclusive public event that is the American census.

Recommendations:

• Increase the number of Area Census Offices in a way that is commensurate with projected increases in the NRFU workload in many hard-to-count communities.
• Open and operate, through the self-response and NRFU phases, Questionnaire Assistance Centers in hard-to-count urban and rural communities (as identified through the Census Bureau’s 2018 planning database) where no ACO will be located.

The 2020 Census Operations

1. Content and Forms Design: the race and Hispanic origin questions

We are deeply disappointed that the Census Bureau was unable to revise the 2020 Census race and Hispanic origin questions to reflect years of painstaking, iterative testing and recommendations from the bureau’s own staff. Those recommendations, in significant part, would have improved the collection of accurate race and ethnicity data overall and for subgroups by combining the race and ethnicity questions and by adding a new reporting category for persons of Middle Eastern and North African (MENA) ancestry. It is troubling that the bureau was unable to move forward with these long-anticipated and important improvements to the race and ethnicity questions due to the Office of Management and Budget’s failure to finalize proposed updates to the Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity (“Standards”), themselves the product of in-depth consultation with the Census Bureau, other federal agencies, and stakeholders. Our organizations, in fact, helped convene stakeholders for meetings and briefings both with expert OMB and Census Bureau staff to provide feedback and recommendations as research and testing proceeded over the decade.

While we understand that the current OMB Standards do not favor the collection of race and Hispanic origin data in a combined question, we nevertheless offer two recommendations to improve the proposed questions set forth in this Notice.

Recommendations:

• Senior Census Bureau and Commerce Department officials should convey to the Director of OMB and the Administrator of the Office of Information and Regulatory Affairs their strong support for revisions to the OMB Standards published in the Federal Register on March 1, 2017 (82 FR 12242), especially with respect to revisions that would allow the Census Bureau to use the combined race and ethnicity question with a discrete MENA category recommended by bureau staff to the Census Director. Bureau and department officials should urge OMB to adopt revised Standards post haste, in order to allow the bureau to use the preferred question, with formatting and instructions that have been researched and tested thoroughly, in the 2020 Census.
• If the Census Bureau is unable to modify the proposed questionnaire to use a combined race and ethnicity question, it must use the Hispanic origin question tested in the 2018 End-to-End Census Test, which includes an instruction to “Mark one or more boxes AND print origins.” There simply is no compelling reason to drop the instructions clearly contemplated and validated by the Census Bureau’s comprehensive testing and evaluated in the dress rehearsal; allowing Hispanics to identify with multiple origins is consistent with the bureau’s goal of improving the measurement of the nation’s diverse population.

For further information on the civil rights community’s perspective, positions, and concerns regarding the collection of race and ethnicity data in the 2020 Census, we are pleased to offer our November 2014 report, “Race and Ethnicity in the 2020 Census: Improving Data to Capture a Multiethnic America,”[6] and comments in response to the March 1, 2017 OMB Federal Register notice referenced above.

1. Language Services

We appreciate the Census Bureau’s efforts to offer response opportunities online and via telephone in 12 non-English languages, an improvement over previous censuses. However, we are not confident that the language support program is sufficiently extensive or robust to facilitate participation for all limited English proficient individuals and households. We are particularly concerned that paper questionnaires will be available only in English and Spanish, as Limited English Proficiency households are far less likely to have access to the Internet, and that census takers will only have access to enumeration materials in English and Spanish (with the exception of identification cards in 59 non-English languages).

We also are concerned that the 59 non-English languages selected for language assistance guides and other support materials include only one American Indian language (Navajo) and no Alaska Native, Native Hawaiian, or Pacific Islander languages. These oversights will make it far more difficult not only for households within these important race categories to participate in the census, but for community advocates to promote census response among populations already at risk of disproportionate undercounting.

Recommendations:

• The Census Bureau should consult post haste with knowledgeable representatives of communities not represented currently in the language support program, to identify ways to offer language assistance during all census operations (e.g., self-response; Nonresponse Follow-up; Enumeration of Transitory Locations and Service-based Enumeration) to limited English proficient American Indian, Alaska Native, Native Hawaiian, and Pacific Islander households and individuals.
• The Census Bureau should seek immediate support from the Commerce Department to secure a waiver from the Office of Personnel Management that will allow the bureau, as in past censuses, to hire work-eligible noncitizens as enumerators, cultural facilitators, and partnership specialists in immigrant communities where language barriers could prevent complete census participation.

Thank you for the opportunity to comment on key components of the 2020 Census plan and to ensure that the voices of the civil and human rights community continue to be heard as the Census Bureau finalizes methods and operations that it will deploy in the coming months and through the enumeration in 2020. If you have any questions about these comments, please contact Terry Ao Minnis, Asian Americans Advancing Justice | AAJC, at 202-296-2300 x0127, or Arturo Vargas, NALEO Educational Fund, at 213-747-7606, or Corrine Yu, The Leadership Conference, at 202-466-5670.

Sincerely,

Vanita Gupta
President and CEO
The Leadership Conference on Civil and Human Rights

Arturo Vargas
Chief Executive Officer
NALEO Educational Fund

John C. Yang
President and Executive Director
Asian Americans Advancing Justice | AAJC

[1] While the 2010 Census Coverage Measurement Survey showed an overcount of many 50+ race and Hispanic origin cohorts, continuing a general pattern seen in previous censuses, we are concerned that an emphasis on Internet response in the 2020 Census could make it more difficult to obtain an accurate count of older Americans, especially in lower income households, who are less likely to regularly use the Internet.

[2] “2020 Census: Actions Needed to Improve In-Field Address Canvassing Operation,” Government Accountability Office, June 2018 (https://www.gao.gov/assets/700/692493.pdf)

[3] http://www.civilrightsdocs.info/pdf/reports/Counting-Everyone-in-the-Digital-Age.pdf

[4] https://www.urban.org/sites/default/files/publication/90446/census_ar_report.pdf

[5] Some communities refer to people who live in abandoned or vacant homes as “squatters.”

[6] http://civilrightsdocs.info/pdf/reports/Census-Report-2014-WEB.pdf