It is difficult to precisely quantify the size of the homeless population. Direct methods of counting the number of individuals experiencing homelessness include a single-night count or census of the homeless where contact is made with each homeless person. An advantage of this approach is that direct contact allows for data accuracy, as well as collection of other sociodemographic and other information on the characteristics of these populations. However, this approach is limited in that it only represents a point-in-time snapshot or cross-sectional account of the state of homelessness at that juncture. In 2017, the national Annual Homeless Assessment Report found that on a single night in 2017, 553,742 individuals experienced homelessness.
The method that yields the largest estimates is surveying people in conventional housing about their past experiences with homelessness. Link et al. (1994) contacted 1,507 adults using random-digit dialing to ask individuals to self-report on their experience with homelessness. Based on this approach it was estimated that 26 million people (14 percent of the nation’s population) had experienced self-defined homelessness during their lifetimes and that 8.5 million people had experienced homelessness in the past 5 years. This included those adults who were “doubled-up”: that is, who moved in with friends or relatives to avoid homelessness. The self-reported group of individuals who had experienced literal homelessness over their lifetimes was 13.5 million people or 7.4 percent of the population. Among this group, 5.7 million (3.1 percent) reported homelessness during the 5 previous years.1
Recognizing these challenges, and partly in response to federal legislation, HUD has worked to create a comprehensive system to count and track persons experiencing homelessness. Local Homelessness Management Information Systems (HMIS) were created and information is aggregated into national reports in response to a Congressional directive in 2001 requiring HUD to provide data to support an assessment of the effectiveness of the McKinney-Vento Act. HUD was
charged with providing accurate and unduplicated counts of the clients using homelessness services and a detailed accounting of the pattern of services used. To meet this charge, from 2001 to 2005, HUD extensively consulted with grassroots organizations who were already working at the community level to use technology to improve service delivery for their local populations. HUD also consulted with technology experts and service providers to collect information on the essential elements of data collection. This planning and needs assessment period resulted in the development of the HMIS Data and Technical Standards guide in 2004. Universal data elements included: name, social security number, date of birth, race and ethnicity, gender, veteran status, presence of disabling conditions, residence prior to use of services, entry and exit date for services, and destination post services. However, each program uses its own computer program for tracking this information, and it is not easy or even possible to share information across jurisdictions.
A 2017 update of the HMIS data guide now requires additional information on the relationships of the user to the head of household as well as the length of time on the street or in an emergency shelter or safe haven prior to receipt of services. There is also a requirement for data collection on program-specific indicators including income and source of income, noncash benefits, presence of physical or development disability, HIV/AIDS status, mental health, substance abuse, domestic violence, services received, destination upon exit and reason for leaving, employment, education, general health status, pregnancy status, veteran’s information, and children’s education.
In another effort to better determine the number of persons experiencing homelessness in America, HUD developed a single point-in-time (PIT) counting system. The origins of this system date back to the early direct methods employed by HUD to count the homeless. The first study was conducted in 1984 in which a subset of service providers was asked to estimate the number of persons experiencing homelessness in their geographic area. Estimates suggested that between 250,000 to 350,000 persons were experiencing homelessness at a given point-in-time. Building on lessons learned from this early work, HUD created a new PIT system that included a mandate requiring Continuums of Care networks to annually count all persons experiencing homelessness in their catchment area including those in emergency shelters, transitional housing, and other safe havens. HUD also standardized the timing of the data collection to a single night in January. More recent estimates based on a single night in January 2017 suggested that 553,742 persons were experiencing homelessness in 2015, for a rate of 17 per 10,000 persons (HUD, 2017a).
Despite improvements made to the PIT system counts, significant logistical challenges remain in this method of data collection. A phalanx of volunteers is deployed to conduct a count of people experiencing homelessness within a specified geographic area; this effort is coordinated by the local Continuum. A variety of concerns have been raised about this procedure, including logistical issues covering large geographic areas and the likelihood of missing individuals experiencing homelessness who choose to remain out of sight and, therefore, are not
counted. Efforts have been made to improve the PIT counts, including using decoy individuals who are “planted” in sites (Hopper et al., 2008) and the use of an incident command system similar to that used by police and fire departments (Troisi et al., 2015). Nonetheless, there continues to be widespread agreement that it is difficult to adequately estimate the number of persons experiencing homelessness in the United States. Although New York City’s count is one of the most sophisticated, Hopper et al. (2008) estimated that it missed about half of the people not staying in shelters.
Thus, there are persistent concerns that estimates based on PIT counts represent a significant underestimate of the true burden of U.S. homelessness. Several sources of evidence provide support for these concerns. For example, the Annual Homelessness Assessment Report provides estimates of the overall number of people who stay in shelters or transitional housing programs in the United States over the course of a year and estimates by specific subgroups; however, the report does not provide information on the number of people at risk of homelessness. One strong predictor of homelessness is “doubling up,” as defined above. HUD’s American Housing Survey found that from 2003 to 2009, the number of doubled-up households of more than one family living together increased from 2,737,000 to 3,354,000 households (PD&R Edge, 2014). Increases were also found for households with an adult child living at home and households with more than one family where individuals are related.
Although “doubling up” is not a form of literal homelessness, HUD (2014) considers it as “housing instability” and thus many researchers see this as a precursor to the experience of homelessness for the family staying with an existing household. Among people who entered homeless shelters from housing in 2016, three quarters had been staying with family or friends and only a quarter in a place they owned or rented prior to becoming homeless (HUD, 2017b). HUD treats doubling up households as experiencing “housing instability” rather than homelessness.
The National Alliance to End Homelessness (2016) estimated that nearly 7 million individuals were “doubled up” in 2014.2 Another household-related factor that serves as a strong predictor of future homelessness is severely high housing costs for low-income renters (Shinn et al., 1998). HUD reports that 7.7 million households, or almost 42 percent of very low income renter household, experience “worst-case housing needs.” Worst-case housing needs are defined as “renters with incomes below 50 percent of the area median income who do not receive government housing assistance and who pay more than one-half of their income for rent.”
2 Most recent data available.
Hopper, K., M. Shinn, E. Laska, M. Meisner, and J. Wanderling. 2008. Estimating numbers of unsheltered homeless people through plant-capture and post-count survey methods. American Journal of Public Health 98(8):1438-1442.
HUD (U.S. Department of Housing and Urban Development). 2017a. The 2017 Annual Homeless Assessment Report (AHAR) to Congress. Part 1: Point-in-time estimates of homelessness. Online. Available at https://www.hudexchange.info/resources/documents/2017-AHAR-Part-1.pdf. Accessed April 6, 2018.
HUD. 2017b. The 2016 Annual Homeless Assessment Report (AHAR) to Congress. Part 2: Estimates of Homelessness in the United States. Online. Available at https://www.hudexchange.info/resources/documents/2016-AHAR-Part-2.pdf. Accessed May 13, 2018.
Link, B. G., E. Susser, A. Stueve, J. Phelan, R. E. Moore and E. Struening. 1994. Lifetime and five-year prevalence of homelessness in the United States. American Journal of Public Health 84(12):1907-1912.
National Alliance to End Homelessness. 2016. The State of Homelessness in America. Online. Available at https://endhomelessness.org/homelessness-in-america/homelessness-statistics/state-of-homelessness-report/. Accessed September 29, 2017.
P&R Edge. 2014. American Housing Survey Reveals Rise in Doubled-Up Households During Recession. Online. Available at https://www.huduser.gov/portal/pdredge/pdr_edge_research_012714.html. Accessed September 29, 2017.
Shinn, M., B. C. Weitzman, D. Stojanovic, J. R. Knickman, L. Jimenez, L. Duchon, S. James, and D. H. Krantz. 1998. Predictors of homelessness among families in New York City: From shelter request to housing stability. American Journal of Public Health 88(11):1651-1657.
Troisi, C. L., R. D’Andrea, G. Grier, and S. Williams. 2015. Enhanced methodologies to enumerate persons experiencing homelessness in a large urban area. Evaluation Review 39(5):480-500.