Count People Where They Are – Center for American Progress

Author’s note: CAP uses “Black” and “African American” interchangeably throughout many of our products. We use “Native American” and “American Indian” interchangeably in this report as many reports referenced include either or both terms in their data collection. We also use the term “Latinx,” which includes a person or group of people with origins in Latin American. This term is preferable to “Latino,” which is not gender-inclusive.

Introduction and summary

In conducting the decennial census, the U.S. Census Bureau endeavors to fulfill a constitutional mandate to enumerate every person residing in the United States.1 The census is a large, vast, and complex operation undertaken by the federal government; while the Census Bureau continually innovates to improve the count, the bureau has historically missed and miscounted certain individuals and households. Given that the census data are used for the apportionment of seats in the U.S. House of Representatives,2 redistricting at all levels of the government,3 and the allocation of more than $1 trillion in federal funds annually,4 among other uses, the undercount of diverse populations deeply undermines the fairness and accuracy of the census and puts undercounted communities at greater political and economic disadvantage. Without accurate representation and funding, many already experiencing multiple hardships may be ignored in efforts to provide crucial and lifesaving services.

People experiencing homelessness have been historically undercounted in the decennial census. For the 2020 census, the bureau has two primary operations for counting people experiencing homelessness:

  • Service-Based Enumeration (SBE): SBE involves census workers counting people at the places they receive services, such as food pantries. SBE also includes counts at Targeted Nonsheltered Outdoor Locations (TNSOL), which involve census workers counting people experiencing unsheltered homelessness.
  • Enumeration at Transitory Locations (ETL): This operation endeavors to count people living at places such as hotels and campgrounds.

People experiencing homelessness may also participate in the census through other operations such as by responding online, by phone, by mail, or through the Group Quarters Advance Contact (GQAC) operation, as well as through the Nonresponse Follow-Up (NRFU) operation. During the NRFU, census enumerators follow up in person with households that did not self-respond.

Due to myriad unprecedented challenges and risks facing the 2020 census, however, based on the bureau’s estimates of omissions, net undercount, and differential undercount rates by demographics and geographical characteristics, people experiencing homelessness may be undercounted and miscounted at higher rates than in previous decades. In particular, key operations such as SBE and ETL have experienced lengthy delays and operational challenges due to the disruptions caused by the COVID-19 pandemic, including truncated timelines for data collection. The pandemic and resulting economic recession make data on homelessness especially important. These crises are expected to make homelessness more common, deepen the challenges faced by people experiencing homelessness, and increase the need for lifesaving public programs. This report looks at the diverse circumstances of people experiencing homelessness, shedding light on how the Census Bureau miscounts them and on some of the federally funded programs that are essential to meet their needs. The bureau must address past and current challenges and take action going forward to more accurately count people experiencing homelessness.

Experiences of homelessness are diverse

People’s experiences with homelessness can be temporary, episodic, or chronic, and living situations can change daily. People experiencing homelessness sometimes live in a street encampment, in a wooded area, in a shelter, in a short-term lease such as a motel, with friends or family temporarily, or in a number of other arrangements.

The Department of Housing and Urban Development (HUD) defines homelessness as lacking regular and adequate shelter meant for human habitation or sleeping at night; living in temporary living arrangements; or exiting a temporary institutional residence.5 The Department of Education, the Violence Against Women Act,6 and the Runaway and Homeless Youth Act7 expand this definition to include families who double up in a single-family unit or leave a home situation to escape abuse. The lack of a clear and consistent definition and standards for measuring homelessness at the local, state, and federal levels makes qualifying the prevalence of homelessness difficult and quantifying it a challenge.

Almost 30 million people are living in inadequate or unhealthy housing,8 such as in a unit without adequate water or electricity or unsafe building conditions. Meanwhile, almost 570,0009 individuals experienced homelessness on any given night in the United States, according to a point-in-time count conducted by continuums of care—local planning bodies responsible for coordinating the funding and delivery of services for people experiencing homelessness. During the 2016-17 school year, 1.4 million children ages 6 to 18 experienced homelessness.10 As the federal moratorium on evictions along with the unemployment assistance established in the Coronavirus Aid, Relief, and Economic Security (CARES) Act expired in July, as many as 28 million to 40 million people were expected to be faced with eviction from their homes in the coming months or year, unless additional measures were taken.11 By comparison, 10 million people were evicted from their homes during and following the 2007–2009 financial crisis.12 In September, the Centers for Disease Control and Prevention declared a halt on residential evictions for households facing financial hardship due to the COVID-19 pandemic.13 However, without rental assistance, many are expected to face eviction after this and local moratoria on evictions do expire. While not everyone who is evicted will face homelessness, evictions are a cause of homelessness—whether a “direct and immediate cause” or one factor that increases a person or family’s risk of future homelessness.14 Estimates are not available for how much homelessness may increase in 2020 based on expected evictions. However, one researcher at Columbia University estimated that the number of people experiencing homelessness in the United States could grow by an estimated 40 percent to 45 percent by the end of the year—an increase of about 250,000 people from prior to the pandemic. The analysis relied on unemployment projections for July 2020 and past impacts of unemployment increases on homelessness; unemployment is also both a cause and consequence of evictions.15

Marginalized groups are likely to be most affected by increased homelessness and therefore a census undercount. Historic and current discriminatory policies and practices have created disparities in education access and attainment, income, access to public services, and other factors that contribute to financial security. People of color,16 LGBTQ people,17 veterans,18 former foster youth,19 formerly incarcerated people,20 people with disabilities,21 families with no and low incomes,22 and other disenfranchised people experience homelessness at higher rates than the general population. For example, according to the National Alliance to End Homelessness, people who are Black, Native American, and Pacific Islander experience significantly higher rates of homelessness than white people.23

The 2010 census undercounted Black people in the United States by 2.06 percent; American Indian and Alaska Natives by 0.15 percent; and Native Hawaiian or Pacific Islanders by 1.02 percent, while white people were overcounted by 0.54 percent.24 People of color are undercounted in decennial censuses for a variety of reasons. For example, Latinx;25 Asian American, Native Hawaiian, and Pacific Islander;26 Black;27 and American Indian and Alaskan Native28 people disproportionately live in hard-to-count census tracks—meaning hard to locate, contact, interview, or persuade.29 Middle Eastern and North African people are categorized as “white”30 and practically not counted at all. Within these groups, many communities of color face continued language barriers,31 distrust of the government,32 and housing instability,33 making accurate counts more difficult. As a consequence, decision-makers may have underestimated the need for assistance, both emergency and long-term, and inadequately resourced crucial programs and services. Moreover, as a result of the pandemic, millions more people may look to assistance programs to meet their basic needs—programs that use decennial census data to determine funding and where to allocate resources regionally and programmatically.

People experiencing homelessness have likely been undercounted in the decennial census for decades

A variety of factors may have contributed to the Census Bureau likely missing and miscounting people experiencing homelessness for decades.34 These undercounts result in unequal political representation and the misallocation of crucial federal and state resources for programs. In particular, programs that serve people experiencing homelessness and their communities, such as housing vouchers and rental assistance programs, help prevent homelessness.

Traditional census enumeration methods may not reach many people experiencing homelessness. The decennial census primarily relies on attempts to count people by reaching out to the addresses of all the housing units and group quarters known to the Census Bureau.35 As a result, these traditional methods are likely to miss people without conventional housing, such as individuals who may be staying in cars, abandoned buildings, or public parks. Furthermore, people experiencing homelessness or who are at risk of homelessness may be inadvertently left out of a household’s census response, as the householder who fills out the form may not think to include them as a part of their household. In 2019, 25.2 million households were estimated to be living doubled up.36 It may not be clear to households with multiple families or relatives, such as adult children, living together temporarily who should be included in the household’s census forms. Consequently, the 2010 census included an item prompt that listed an option to include nonrelatives and people residing in the household temporarily,37 and the 2020 census forms have included a prompt asking respondents about additional people in the household, nudging them to include people who are staying for a short or indefinite period of time by asking if the person usually lives or stays somewhere else.38

In an effort to improve the count of people missed by traditional census methods, the bureau has relied on special enumeration operations such as SBE, TNSOL—a suboperation of SBE—and ETL. Broadly, these operations seek to count a portion of the population experiencing homelessness by locating them at places where they receive services, such as shelters and meal centers, and at outdoor locations, as well as by counting individuals at transitory locations, such as campgrounds, motels, and marinas.39 The 2010 census counted almost 422,972 people through the SBE, including TNSOL, operation.40 (see Figure 1) For the 2010 census, the bureau combined the count of people enumerated through the ETL operations with the count of people living in housing units; a separate figure of the number of people counted at transitory locations is not available.41 The 2020 census will be the second decennial census in which ETL is conducted separately from the Group Quarters Operation.42

Figure 1 

While the bureau has made improvements since the introduction of the first nationwide effort in the 1990 census to count people at emergency and transitional shelters and certain outdoor locations,43 fundamental challenges remain. Potentially as a consequence of the criminalization of homelessness and police harassment,44 some people experiencing homelessness may remain distrustful of and avoid government workers—including census takers—making a complete and accurate count more difficult. Others may choose not to participate in the census due to concerns that their responses may adversely affect their access to public services and benefits.45 When deciding whether to participate in the census, people may not be aware that the Census Bureau is bound by law to keep census responses confidential.46 Within its operations, it is important that the Census Bureau strives for transparency, cultural competency, and trust with people living in various circumstances and with different needs.

Additionally, the bureau’s point-in-time methodology provides only a limited snapshot of the number of people served at the various service, outdoor, and transitory locations. The bureau may not exhaustively identify all possible locations where people experiencing homelessness are staying, and not everyone at the identified locations may be counted or counted accurately. Even when counted, a percentage of people experiencing homelessness may not be “data-defined”—meaning they may have only one or zero characteristics recorded. In particular, unsheltered people are more likely to not be data-defined: In the 2010 census, 13.5 percent of people counted at outdoor locations were not data-defined, compared with 1.3 percent at shelters and 2.4 percent at meal centers.47 When people are not data-defined, it can be more difficult to assess disparities by various characteristics and may undermine the perceived need for programs promoting equity.

Current challenges and operational concerns

As a result of uncertainty about transmission risk associated with COVID-19, the Census Bureau delayed by six months in-person operations that are key to counting people experiencing homelessness. Due to a number of factors, the bureau has also shortened the duration of some key operations. (see Table 1) The timing of these operations will make full and competent staffing for these operations difficult. The bureau will also need to count more people using these operations than anticipated as a result of the economic impacts of the pandemic.

Table 1

The landscape of SBE and TNSOL has shifted

In February 2020, the bureau launched its GQAC operation, during which it contacted all known group quarters facilities to “explain the enumeration process and collect certain information about their group quarters.”48 The SBE and TNSOL operations are a subset of the Census Bureau’s Group Quarters Operation49—which also includes GQAC, which prepares for enumeration; Group Quarters Enumeration, which includes all group quarters such as nursing homes and jails and prisons; and Maritime and Military Vessel Enumerations—through which the bureau attempts to enumerate all people living in congregate settings, such as residents of nursing homes, students living on college campuses, and people living in military barracks, prisons, and jails.

In-person census operations were suspended in March, before the originally scheduled SBE and TNSOL operation occurred. Since that time, the landscape of services and supports for people experiencing homelessness has changed dramatically. Many homeless shelters have closed completely, while others have closed to new clients or have significantly reduced the services offered on-site.50 At the same time, new temporary shelters and services have opened in areas that have never before offered coordinated care as communities recognize the heightened health risk for people experiencing homelessness in their regions.51

Aware that the SBE and TNSOL landscape has shifted since February, the Census Bureau is conducting a second round of GQAC operations in advance of the new group quarters enumeration dates.52 Unfortunately, reports from census and housing advocates and city census leaders across the country indicate that this round of GQAC is being conducted inconsistently and that individuals are receiving conflicting information about how SBE and TNSOL operations will be conducted in their area.53 For example, it was revealed in late September 2020 by the House Oversight Committee54 that on September 3, 2020, just before the onset of SBE and TNSOL enumeration, census employees were instructed to “[o]nly enumerate the locations in your workload. The deadline for adding potential SBE locations to the workload has already passed.” This conflicted with the guidance given to advocates that stated that SBE locations could be added as late as September 21, 2020.55

Recently, Census Bureau staff have made clear that the responsibility for adding new or temporary shelters or services—such as the scores of shelters and food pantries that have opened since the start of COVID-19—to the GQAC workload rests with service providers. As a result of low staffing capacity at many of these shelters and services, proactively reaching out to the appropriate staff person at the bureau is likely a low priority; many new or temporary shelters that have not participated in a prior decennial census may also be unaware that the Census Bureau expects them to make contact with the bureau in order to be included. Because of the miscommunications about the operational plan and the passive structure of the GQAC operation in 2020, it is likely that the renewed GQAC operation is successfully eliminating closed shelters from its workload but is not adequately connecting with new shelters and services not in the original GQAC universe.

Staffing concerns

The Census Bureau’s current operational plan for SBE and TNSOL calls for hiring a dedicated workforce of approximately 45,000 enumerators focused on these operations.56 Early reports from geographies with NRFU operations have begun to indicate that the bureau is having difficulty recruiting, hiring, and retaining staff for the standard enumeration operations.57 If that trend continues, the bureau may not be able to meet its goal of hiring an additional 45,000 enumerators for the SBE and TNSOL operations.

Even if the bureau successfully meets its hiring and retention goals, staffing may be insufficient to conduct a robust SBE and TNSOL operation given the rapidly expanding population of people experiencing homelessness and housing instability. In 2010, the Census Bureau approved hiring up to 60,000 enumerators for this operation; the authorization for this number was based on a higher-than-expected number of TNSOL locations identified in the GQAC operation.58 Despite an expanded workload, however, the number of enumerators proposed for this operation has decreased by 15,000 for the 2020 census. While technological improvements such as efficiency in route mapping and mobile internet response capabilities have been cited as a reason for decreases in overall staffing for the 2020 census,59 it is unlikely that those improvements will be adequately consequential to the SBE and TNSOL operation.60 Moreover, given that the bureau is understaffed across operations,61 it may need to devote less time to training and quality check activities, both of which would undermine the efficacy of the operation and the quality of resulting data.

Reduced support from community partners

In 2010, the Census Bureau relied heavily on its community partners to assist in identifying where encampments were located, to identify language skills needed for these operations, and to function as cultural facilitators.62 The bureau provides guidance for community-based organizations looking to engage in get out the count campaigns that includes a toolkit that highlights best practices for outreach, cybersecurity, media, accessibility, and more.63 Since 2010, the bureau’s Partnerships Program has worked to maintain and expand its partnerships network, working with thousands of advocacy organizations, service providers, and businesses across the country to build support for the 2020 census—including many organizations and providers that serve people experiencing homelessness and housing instability. Had the SBE and TNSOL operation been conducted on its original timeline in March and April 2020, it is likely that many of the organizations involved in the bureau’s Partnerships Program would have played a role in the SBE and TNSOL operation that mirrored their involvement in 2010. Unfortunately, the delayed timeline and the presence of COVID-19 likely makes that level of involvement more difficult for many organizations.

First, many service providers have reduced capacity due to the pandemic: Not only is there a higher need for services, but many providers report a significant reduction in staffing and volunteers due to quarantine or social isolation.64 In a recent study for continuums of care that reported staffing data for homelessness assistance programs, 60 percent reported staff shortages, 88 percent reported shortages in front-line shelter workers, 58 percent reported shortages in street outreach workers, 63 percent reported volunteer shortages, and 46 percent reported shortages in social workers.65 Because of their reduced capacity, conducting additional activities outside of their core services—including census outreach—is likely a more difficult proposition. Second, the landscape of organizations providing services and support to people experiencing homelessness has shifted significantly since the start of the year, so the Partnerships Program may not have relationships with many organizations that are equipped to support the SBE and TNSOL count.66 Finally, because of the forementioned capacity shortages, the Census Bureau is currently off-boarding partnerships specialists—bureau employees responsible for building relationships with partnering organizations. As this off-boarding occurs across the country, including in areas with high rates of homelessness and housing instability, it may be difficult for the bureau to manage the workload of SBE and TNSOL without the relationships built by the Partnerships Program.

Mitigation strategies are less accessible during the pandemic

All these gaps in the Census Bureau’s operational plans for SBE and TNSOL could be mitigated by a strong push for people experiencing homelessness to self-respond to the census. However, that mitigation strategy is more difficult in 2020 due to operational changes to the census and an increased reliance on the internet response option. In 2010, the Census Bureau made blank census “be counted” forms available to people in public spaces such as libraries, community centers, and census questionnaire assistance centers.67 In 2020, the Census Bureau chose not to make these blank forms available to people, instead relying on a Mobile Questionnaire Assistance (MQA) model that would make internet response available to people at public events that draw a large audience of historically undercounted people.68 When the bureau suspended in-person activities, the MQA program was also suspended.69 At the same time, many libraries,70 community centers,71 and other public spaces closed their doors, leaving many people experiencing homelessness without reliable internet access. Though many advocates are working to conduct outreach to people experiencing homelessness to supplement the Census Bureau’s operational plans, the lack of blank census response forms and increased reliance on the internet response option makes that supplemental mitigation strategy more difficult.

Missed opportunity to enumerate marginally housed people through SBE and TNSOL

The operational concerns identified above may imply that SBE and TNSOL are primarily intended to count people experiencing homelessness. However, these operations have the potential to capture many people who are marginally housed as well. Researchers focused on the count of young children in the census have found it difficult to convince people to include unrelated or temporary household members on their census forms.72 In 2019, about 20 percent, or 25.2 million households,73 were estimated to be living doubled up due to economic insecurity, a rise from previous years. In 2020, doubling-up rates have risen during the pandemic and recession.74 Many people who are living doubled up because of economic insecurity75 receive services from providers in the SBE universe. The SBE operation therefore has the potential to capture people missed in the broader NRFU operations if the Census Bureau takes a more robust approach to the operation and accounts for changes in household living experiences over the course of the COVID-19 pandemic.

Census data affects federal funding for lifesaving programs for many groups of people

More than 300 federal programs that allocate funds based on census data affect individuals and families who experience or are at risk of homelessness across circumstances and their intersections. Though program support is affected by various practices and actions, undercounts and miscounts of households and people experiencing homelessness contribute to underestimates of the need for programs that benefit millions and could help people maintain a decent standard of living. The Supplemental Nutrition Assistance Program (SNAP), for example, helped 3.2 people achieve a standard of living above the poverty line in 2018,76 while Supplemental Security Income (SSI) helped 2.9 million people, and rent subsidies helped 3 million people.77 These programs help to stabilize the lives of many people experiencing or at risk of experiencing homelessness. Accurate census counts of people experiencing homelessness will not only provide better data to combat the current crises but will also facilitate more robust preparation for future crises.

1 2