The Demographics of Homelessness - The Authoritative Estimates

Broad national assessments of homelessness were undertaken by several agencies and organizations during the 1980s and mid-1990s, including A Report to the Secretary on the Homeless and Emergency Shelters (Washington, DC: U.S. Department of Housing and Urban Development, 1984), America's Homeless: Numbers, Characteristics, and Programs that Serve Them (Martha Burt and Barbara Cohen, Washington, DC: Urban Institute, 1989), and Homelessness: Programs and the People They Serve, Findings of the National Survey of Homeless Assistance Providers and Clients, (Martha Burt et al., Washington, DC: Urban Institute Report to the Interagency Council on the Homeless, December 1999). In 2002 Burt and other researchers summarized the difficulty of addressing homelessness without a continuing census or other governmental program to track the homeless population in Evaluation of Continuums of Care for Homeless People (Washington, DC: U.S. Department of Housing and Urban Development, May 2002). The report notes that:

Basically, there are only three sources or original data on which to base estimates of incidence (the number of people homeless on a single day) for the nation as a whole—HUD's 1984 effort (HUD, 1984), the Urban Institute's 1987 study (Burt and Cohen, 1989), and the 1996 National Survey of Homeless Providers and Clients (Burt, Aron, and Lee, 2001). Any national estimates offered by anyone for any years other than 1984, 1987, and 1996 are projections or manipulations of one of these three data sources, and include assumptions of population change or growth that are not grounded in data. HUD's 1984 study was based on a survey of providers, who supplied their best guesses as to the size of the homeless population in their cities. Only the 1987 and 1996 studies are based on statistically reliable samples of homeless people using homeless assistance programs. Using these three data sources, the number of people homeless at any one time appears to have grown substantially from the mid-1980s until the mid-1990s—from 250,000-350,000 in 1984 (HUD's "most reliable range") to 500,000-600,000 in 1987, to 640,000-840,000 in 1996. Best guesses or projections of the number of people homeless during the course of a year come from various sources.… These estimates, using different approaches, nevertheless converge on figures that between 2.5 and 3.5 million people (including children) experience at least one night of homelessness within a given year.

Even these data, considered by the government to be reliable, are based on very small samples. The 1996 data, the most recent and most widely used, were based on interviews with 6,300 homeless program representatives, held in February 1996, and interviews with 4,200 users of homeless programs conducted in October 1996. The total number of people homeless at some point in the year 1996 was derived by projection from this sample. While such methods of estimating are common in statistical analysis, they also show that current knowledge about homelessness is, at best, partial.