by
Martha R. Burt, Ph.D.
This paper summarizes the latest and/or most comprehensive data on important characteristics of homeless people. It looks at the demographics and distribution of homeless people among communities of different types, as documented by a range of research methodologies in various jurisdictions and nationwide. It also examines how characteristics may differ depending on the locations in which a study looked for people to include, and factors that seem to make people vulnerable to homelessness.
The paper then turns to the need of local jurisdictions for information to help with service planning. It discusses the variety of people and agencies that might need information for planning, the types of decisions they must make, and what types of information would help them the most. It continues with a review of several strategies that work at the local level for collecting the most useful data, and the advantages and disadvantages of each method. Finally it draws the conclusion that every jurisdiction will be best served by gathering its own information about service needs for planning purposes.
Lessons for Practitioners, Policy Makers, and Researchers
This paper starts with a summary of the latest and/or most comprehensive data on important characteristics of homeless people. It looks at the demographics and the distribution of homeless people (at a single point in time) among communities of different types. It also examines how characteristics may differ depending on the locality where people are found, and factors that seem to make people vulnerable to homelessness.
It then turns to the issue of what might be meant by “need”. It discusses the variety of people who use information on needs among homeless people to make planning decisions, the types of decisions they must make, and what types of information might help them the most. The paper concludes by reviewing several strategies for obtaining data at the state and local levels, and the advantages and disadvantages of each for the various decision makers.
Demographics and Patterns of Homelessness
Many studies have collected descriptive information about homeless populations over the past two decades. Most are studies of particular cities or parts of cities, and some analyze only the information from people staying in a single shelter. Some have specialized purposes such as examining the nature and extent of mental illness or substance abuse or the situations of homeless families, while others are quite general. No attempt has been made to summarize all of these studies. Rather, several of the most recent studies that have methodological interest, cover sizeable geographical areas, and provide overviews of homeless populations are reviewed.
Table 1 summarizes these studies, which include one that covers the entire United States (1990 Census S-Night), three of specific cities (New York, Philadelphia) or parts of cities (Los Angeles), one that is representative of all cities over 100,000 in the United States, one that covers an entire major Metropolitan Statistical Area (Washington, DC), two that provide important new information on homelessness in rural areas (Ohio and Kentucky), and one that summarizes studies on family homelessness. To help in interpreting the basic demographic information shown in Table 1, the table also includes the year(s) in which the studies were done, the types of venues where the studies located their respondents, and the methodological approach used. Gender (percent male), race/ethnicity, education (percent high school graduate or more education), whom respondents are with, and length or patterns of homelessness are the demographic and descriptive data examined in Table 1.
TABLE 1 |
|||||||||
| Study | Census | Urban Institute | Los Angeles COH study | New York, Culhane** | Philadelphia, Culhane** |
DC*MADS | Ohio | Kentucky | Rossi, Family Homelessness |
| Geographic Coverage | United States | US cities of 100,000+ (178 cities) | Los Angeles, downtown and west side | New York City | Philadelphia | Washington, DC metropolitan area (DC, 10 counties, 5 VA cities) | rural Ohio counties | Kentucky
(heavily rural) |
various |
| Date | 1990 | 1987 | 1991 | 1990-92 | 1990-92 | 1991 | 1990 | 1993 | 1985-91 |
| Data Collection Venues | SH | SH, SK | SH, SK, ST | SH | SH | SH, SK, ST | SH, SK, O | SH, ST, O | various |
| Methodological Approach | census | randomly selected cities, programs within cities, people within programs | sample allocated proportion-ally to locations, then random; longitudinal | shelter tracking database | shelter tracking database | combination of block probability and service-based, plus encampments | snowball | service-based plus outdoor search | various |
| Gender of Adults--% male | 70 | 81 | 83 | 52/59 | 45/59 | 76 | 49 | 68-urban
39-rural |
0-27 |
| Race/Ethnicity
%African-American %White % Hispanic % Other |
41 49 16* 10 |
41 46 10 3 |
58 21 14 8 |
65/65 5/8 29/24 1/3 |
91/88 6/8 3/3 0/1 |
76 17 6 -- |
10 85 3 2 |
12 85 1 2 |
10-91 4-85 2-33 -- |
| Education—% high school graduate or more education | NA | 52 | 62 | NA | NA | 60 | 58 | NA | NA |
| Alone/With Others
% Men by themselves %Women by themselves %Female-headed with children %2-parent with children %Other |
NA |
75
|
70
|
45/56
|
|
|
|
|
-- --
|
| Length/patterns of homelessness | NA | Mn=39 mo; Md=10 mo;
21% LE 3 months; 41% GE 2 years |
45% GT 1 year; 67% homeless 2+ times | 78% transitional;
12% episodic; 10% chronic |
81% transitional;
9% episodic; 10% chronic |
18% 1st time;
39% intermittent; 23% chronic; 13% former; 7% never |
Mn=221 days;
50% LE 49 days; 6% GT 2 years |
64% 1st time and LT 1 mo;
10% multiple episodes and GT 12 mo |
Md=LT 1 mo-3.5 mo |
| SH=Shelters,
SK = soup kitchens/meal programs, ST = "streets," O = Other. * Census asks
Hispanic status and racial status separately, so the categories are not mutually
exclusive. ** First number is average daily (point-in-time); second number
is cumulative and unduplicated for the period June 1990 through May
1992.
Sources: Urban Institute—Burt & Cohen, 1989; Census—Barrett, Anolik & Abramson, 1992; Ohio—First, Rife & Toomey, 1994 and personal communication; DC*MADS—Bray, Dennis & Lambert, 1993 and personal communication; LA COH—Koegel, Melamid & Burnam, 1995 and personal communication; Kentucky—Kentucky Housing Corporation, 1993; Rossi, Troubling Families, 1994; New York & Philadelphia—Culhane et al., 1994; Kuhn & Culhane, 1998. |
|||||||||
Looking first at gender, males as a proportion of the homeless population range from a high of 83 percent (in Los Angeles) to a low of 39 percent (rural Kentucky), excluding for the moment a consideration of the family homelessness studies reviewed by Rossi. The statistics on gender from the Table 1 studies lead to the generalization that the more urban/central city a place is, the higher the proportion male among the homeless population. Conversely, the more suburban/rural a place is, the higher the proportion female, largely due to the higher ratio of families to singles in the suburban/rural jurisdictions. New York and Philadelphia appear to be the exceptions to this generalization, probably because New York unlike most cities really does have a high ratio of families among its homeless population even at single points in time, and the Philadelphia database probably misses many of the single men who use the biggest shelter. The studies reviewed by Rossi, concerned as they are exclusively with family homelessness, report much lower proportions of adults who are male.
Race/ethnicity varies considerably among the Table 1 studies, owing largely to the variation in the racial and ethnic composition of the communities where the studies were conducted. Regardless of location, however, African-Americans are significantly overrepresented among homeless people compared to the general population. Compared to 12 percent of the U.S. population who are African-American, the Urban Institute's 1987 study found that 41 percent of homeless people in large U.S. cities were African-American (and not Hispanic) while the 1990 S-Night counts in shelters found that 41 percent of the people enumerated were African-American (including those who were also Hispanic).
People of Hispanic origin do not appear to be consistently overrepresented among homeless populations. For instance, DC*MADS found 5.9 percent of the homeless population to be Hispanic compared to 5.2 percent of the total 1990 population of the Washington, D.C. MSA, and Culhane and colleagues (Culhane, et al., 1994) report that 3.4 and 27.2 percent, respectively, of Philadelphia and New York shelter users over a three-year period were Hispanic, compared to 1990 Hispanic populations of 5.6 percent in Philadelphia and 24.4 percent in New York City.
There is substantial agreement among the studies as to the educational achievement of respondents, with 52-62 percent having completed high school or a higher level of schooling. With respect to whether people were homeless by themselves, with children, or in some other arrangement, studies differ but there is some systematic variation we can account for. Urban studies that went beyond shelters to include substantial parts of the street population found that single men comprised three-quarters or more of the "households." The more rural the location of the study, the larger the share of households that are families with children. Also, when one relies on shelter data only, as in the case of the New York and Philadelphia tracking data bases, the proportion male declines and the proportion of families with children (most headed by females) goes up. When a data source leaves out the relatively large proportion of single homeless men who do not use shelters or, as in the case of Philadelphia, does not count the shelter use of the most erratically shelter-using single men, the omission distorts the picture of household structure among homeless people.
Homeless youth are one part of the homeless population often missing from policy consideration. Most studies of homeless populations do not include a significant number of youth homeless on their own, both because most of the venues where studies go to find homeless people serve only adults, because many homeless youths are reluctant to use services at all, and because it is difficult to identify homeless youth with other study techniques. Estimates of youth homelessness are often given not as point-in-time estimates but as "homeless within the past 12 months" estimates, and range from about half a million to a million and a half (Ringwalt et al., 1998). Ringwalt et al. (1998) used recent data from youth interviewed from households living in conventional dwellings for the Youth Risk Behavior Survey to estimate that 1.6 million youth ages 12 to 17 (7.6 percent [+ 0.7 percent] of the population in this age range) had a homeless episode of at least one night's duration within the 12 months before being interviewed. Only about 2 in 5 of these youth said they used shelter services during their homeless episode.
This review makes clear the great extent to which basic information about homeless people varies by geography and also depends on the venues from which people are interviewed or data are assembled. It is very important for decision makers to be fully aware of the inclusions and omissions in the data they use for planning, because different data sources (e.g., shelter only, shelter and other services, or services plus "street" sources) may lead to quite different assessments of population characteristics and hence of need.
The studies in Table 1 report their sample members' length of homelessness or patterns of homelessness quite differently. Some report the length of time that people have been homeless during their current spell, some divide their sample into groups such as chronic, episodic, and transitional/crisis/first time. Findings vary considerably, but some generalization may be possible. However, to interpret the length/patterns data correctly, it is necessary to leave out the New York/Philadelphia tracking database information for the moment. Looking first only at the point-in-time data, it appears that we can say four things: (1) long spells (more than 1 year) and or chronic homelessness characterize urban samples much more than they characterize rural areas; (2) in urban areas about 40 percent or more report long current spells; (3) more people in the rural than in the urban samples are in their first spell of homelessness, and these are quite short; and; (4) the higher the proportion of families in the study, the shorter the spell length or the more short spells are reported.
Looking next at the New York and Philadelphia data, it is clear that these multi-year unduplicated data show quite the reverse of these generalizations. Indeed, the vast majority of the homeless population (using shelters) in these cities are transitional (first-time or short-term). With the advantage of unduplicated data covering several years, the New York and Philadelphia results make very clear the dangers of relying on point-in-time data to describe what proportions of the homeless population have spells of different lengths or different patterns of homelessness. Point-in-time data will always be biased toward showing higher proportions of longer spells. When planners use point-in-time data, which they most often will because that is all they can get, it is very important to try to compensate for its biases toward long spells and away from short ones. If this is not done, the whole system of homeless services may be structured in ways that are not in tune with the needs of the people coming for assistance.
Quite a number of studies, both longitudinal and cross-sectional with comparison data, have documented strong associations of negative childhood experiences with homelessness (Bassuk, et al., 1997; Caton et al., 1994; Herman et al., 1997; Koegel, Melamid and Burnam, 1995; Mangine, Royse and Wiehe, 1990; Susser, Struening and Conover, 1987; Susser et al., 1991; Weitzman, Knickman and Shinn, 1992; Wood et al., 1990). The most common childhood experiences associated with a higher risk of experiencing homelessness are: histories of foster care and other out-of-home placement, physical and sexual abuse (which often precede out-of-home placement), parental substance abuse, and residential instability and homelessness with one's family as a child. These experiences are much more common among people who have been homeless as adults than among people who have not.
In addition to evidence that risk factors from an individual's childhood predispose to homelessness, an important new type of research documents the contribution of certain environments to homelessness even after taking into account the characteristics and histories of those who become homeless while living in these environments. Culhane, Lee and Wachter (1996) analyze the addresses of families applying for emergency shelter in New York City and Philadelphia, and find them much more concentrated geographically than poverty in general. (Culhane has recently obtained similar results for the neighborhoods of origin of homeless families in Washington, DC.) Neighborhoods producing high levels of family homelessness have high concentrations of poor African-American and Hispanic female-headed households that include children under six years of age. The housing is the poorest in the city, and despite the fact that rents are the lowest available, residents still cannot afford them, with the consequence that housing is overcrowded and many families double up, even though apartment vacancy rates are high. These conditions create a large pool of families at risk of homelessness, from which it only takes a small percentage every week to fill the available homeless shelters.
The matter of where homeless people may be found can have several interpretations. The first examined here is: how are homeless people distributed geographically among central cities, the suburbs and urban fringe areas that make up the balance of the territory within Metropolitan Statistical Areas (MSAs), and rural areas (outside of MSAs)? The second is how homeless people are distributed among types of services and street locations within communities.
Only a few studies exist that can shed light on the issue of geographical distribution because only a few studies include any geographical diversity in their sampling. The 1990 Census counts of people in emergency, domestic violence, and youth shelters on the night of March 20-21, 1990, which was done on a single night and treats everyone, including children, as individuals, found 75 percent in central cities, 18 percent in suburbs and urban fringe areas (the parts of MSAs that are not central cities), and 7 percent in rural areas (Burt et al., 1993Cthese figures do not include anyone counted in the "visible in the streets" part of the 1990 Census). This compares to the overall 1990 U.S. population distribution of 32 percent in central cities, 43 percent in suburbs and urban fringe areas, and 25 percent in rural areas.
In Kentucky's 1993 study, one of the few that covers a whole state and goes well beyond shelters), 21 percent of homeless individuals were found in the three major urban areas of Louisville, Lexington, and Covington which have 25 percent of the state's 1990 population, while 79 percent were found in the remaining 117 counties of the state where 75 percent of the population reside.(1)
In DC*MADS, Washington, D.C. accounted for 77 percent of homeless and transient individuals, with the remainder found in the suburbs and urban fringe (no areas were included in the study from outside the MSA). In the Washington, DC MSA as a whole, only 15 percent of the population is located in Washington, DC.
Everyone expects to find more homeless people in highly urbanized areas than in suburban and rural areas, but it is also true that the rate of homelessness per 10,000 people has been shown to vary considerably even when one considers only homelessness in large cities. DC*MADS estimates produce a homelessness rate of about 150/10,000 population for Washington, DC and a rate of about 33/10,000 for the whole DC metropolitan area. Using Shelter Partnership's estimate of 80,000 to 90,000 homeless people in Los Angeles County produces a homelessness rate of 88/10,000 to 99/10,000 for the county as a whole in the late 1990s. Burt (1992a) obtained rates for U.S. cities with 100,000 or more population in 1986 ranging from under 10/10,000 up to 65/10,000 (average=18/10,000) based on the number of shelter beds available in the city rather than on actual estimates of homeless people. This magnitude of variation strongly suggests the wisdom of having one's own local data rather than relying on national averages.
A second way to think about "where" homeless people are is to think about the services they might use, and therefore where they might be found within a community. Important locations where people might be found include streets, outdoor locations, and other locations “not meant for human habitation” (for short, “the streets”). Although the streets are not a service, one important issue for planning is how many homeless people are being missed if one focuses one’s data collection efforts only on those who use services, and what types of services they might need. This was a very serious problem when studies went only to shelters, or attempted to augment shelter-based data collection with a street component, because street searches are almost always unsatisfactory. Often they do not locate many people, and they become more dangerous to do the more thoroughly one tries to get to the most hidden places. So the problem is, what is a safe and comprehensive way to include homeless people who do not use shelter services in data collection?
A breakthrough, not less helpful because it was serendipitous, occurred in our ability to include a large part of the non-shelter-using homeless population in data collection when studies began including soup kitchens and other feeding programs in their service samples (1987 Urban Institute study, DC*MADS). As we learned when this happened, many currently and formerly homeless people who do not use shelters do come to soup kitchens. The 1996 National Survey of Homeless Assistance Providers and Clients (NSHAPCCTourkin and Hubble, 1997) extended this concept even further to include a wide variety of homeless assistance programs, and the Kentucky Housing Corporation went well beyond homeless assistance programs in its efforts to locate homeless people.
Results from several of these studies are telling. In the 1987 Urban Institute study, 36 percent of homeless adults and 22 percent of children in homeless families used both shelters and soup kitchens in the week before being interviewed. Thirty-two percent of homeless adults and 73 percent of children in homeless families only used shelters, and 29 percent of homeless adults (but only 5 percent of children in homeless families) used only soup kitchens in the past week (Burt and Cohen, 1989, p. 37). Thus inclusion of soup kitchens in the study design increased the coverage of non-shelter users considerably.
DC*MADS added a soup kitchen component to its design after finding very few people in the street part of its original shelter/street design. The resulting ability of DC*MADS to map the overlapping movements of its respondents produced very interesting patterns. Fifty-six percent of respondents had used a shelter within the previous 24 hours, 65 percent had used a soup kitchen, and 21 percent had spent time on the streets. The overlap of these populations was considerable, with 27 percent using both shelters and soup kitchens. Of all the respondents to DC*MADS, only 7 percent would not have been found if the study had left the street component out entirely and gone only to shelters and soup kitchens (Bray, Dennis and Lambert, 1993, p. 3-3). This was true for the entire literally homeless population in DC*MADS, as well as for the population including transients.
The degree of population coverage achieved by DC*MADS through its shelter and soup kitchen components is very encouraging, but must be qualified in several ways. Even within DC*MADS, subgroup analysis revealed that coverage was somewhat worse without the street component for some groups. The people least likely to be captured were heavy alcohol users, about 15 percent of whom would have been missed without the street component. On the other hand, coverage for those with drug use during the month of the survey was actually better than for the sample as a whole (over 95 percent). Other evidence for differential coverage by population subgroup comes from the Course of Homelessness study in Los Angeles, where only 16 percent of young, single men on the west side of Los Angeles would have been captured with a design that relied only on shelters and soup kitchens, without a street component (Koegel, personal communication, 1996).
A final caveat is that DC*MADS achieved its level of coverage using a homeless-specific service-based approach in an environment where many homeless-specific services are available. In environments such as many suburbs and rural areas where homeless-specific services are scarce or nonexistent, such an approach would clearly miss almost everyone. In these environments more and different types of service agencies would have to be incorporated into the design to achieve adequate coverage, as was done in NSHAPC and in the Kentucky statewide study. Even then, the Kentucky Housing Corporation augmented its service-based approach with a targeted search of outdoor locations.
Many of the studies reviewed in Table 1 used quite sophisticated methodologies, and produced elegant and reliable results. However, they were very expensive to conduct and, while they may be helpful at the national level and to answer particular research questions, are of limited utility to local planners. The most important lesson to be learned from these studies is that even expensive, methodologically sophisticated studies cannot produce consistent findings because the reality of homelessness varies a good deal with the geographic location of interest. Therefore, local decision makers should make every effort to collect their own data using less perfect but a good-enough method, collect it with sufficient regularity and thoroughness that it becomes a useful tool for decision making..
Information for Planning: Who Needs What?
Many people may be involved in planning homeless service systems or in estimating how much service is needed at a given time. Table 2 shows the variety of people who plan, from administrators in direct service programs up though the staff of federal government agencies. It also shows the things they may plan for (column 2) and the information that might help them accomplish this planning, together with the sources that might provide the information (column 3).
Planning may be very simple, such as predicting how many meals to prepare, how many cots or mats will be needed for overflow conditions, or how many nurses will be needed for a health clinic at Shelter X on the next Tuesday. These types of decisions are very local and very practical. Usually they are made at the program level on the basis of past experience, without a great deal of data-based analysis except perhaps to look at agency records of services delivered, if they exist. Temporal variations over the week, month, or season are also important for planners at this level, and will most likely be based on historic agency data or personal experience.
A similar type of planning may occur at the city or county level as officials try to anticipate how many shelter or transitional housing beds might be required to accommodate average and maximum demand, and how many services of other types might be appropriate for shelter users. The simplest approach to this is to ask what was used last year. Need for growth or change in the system's capacity could be approached by examining local economic factors (e.g., plant closings, economic downturns), and possibly also by looking at service requests that could not be met anywhere in the system as an indicator of unmet need. Care must be taken, however, to be sure that one does not count persons turned away from one facility on one day who receive the requested services either from another provider on the same day or any provider within a few days of the request.
Once planning advances to questions of the types of service that ought to be available, to planning a comprehensive and accessible continuum of care, or to client length of stay and its relation to needs for different types of services at different points during a stay, more detailed data are needed if planning decisions are to be driven by facts. It will probably be important to be able to anticipate client characteristics that call for different program structures and services.
For instance, knowing the proportion of households with children who will ask for homeless assistance, in comparison to women or men by themselves, may help agencies or whole communities structure their emergency and transitional shelter resources to accommodate these different types of households. Likewise knowing that, historically, half the people asking for help have some type of immediate health problem may let agencies or community planners prepare to treat those problems. Knowing how many people using homeless assistance programs suffer from mental illness, chemical dependency, and other debilitating conditions can indicate what the level of need for those services are in the population. Finally, knowing how many people are released from psychiatric hospitalizations, detoxification programs, jails or prisons without any reliable plans for housing can provide clues about the demands these people will make on the emergency services system, and possibly also provide documentation to support enhancing the capacities of mainstream systems to take responsibility for aftercare so as to prevent homelessness among these vulnerable groups.
An important issue for planning is knowing how clients see their needs, and how their perceptions might differ from the ways that agency staff see needs. Clients tend to focus on the end point (a job, an apartment), while staff tend to focus on the steps that need to be taken before that endpoint can be achieved (gaining skills, conquering addictions). Both are important. It may not always be easy for staff and clients to reach a meeting of minds about what needs to be done today and tomorrow if the clients' ultimate goals are to be reached. On the other hand, planners must not lose sight of the need for more jobs, more housing, and more services that help people keep their jobs and housing. More case management will not help people get jobs and housing if there are no jobs or housing available.
Another significant issue for planning is knowing whether the clients coming into homeless assistance programs will be short-term or long-term users. This is especially important in shelter/ housing and health programs, where the intake and other procedures that occur on the first day are often the most costly and absorbing of staff time. Client flow is also important in planning caseworker load and types of services to offer. If a program's clients or a whole city's homeless population are mostly short-term, a greater proportion of resources will need to be dedicated to intake than if most of the clients are long-term.
| TABLE 2
INFORMATION NEEDS FOR PLANNING |
|||
| Who Plans? | What Do They Plan For? | What Information Might They Need? | Where Might They Get the Information? |
|
|
|
|
In addition, some services will not be appropriate or needed for short-term clients. One should not plan to put every client through an assessment process that takes three weeks if the average length of stay is two weeks, nor can one expect to produce major life changes in two weeks. What is being asked for is, quite literally, emergency assistance. With respect to transitional programs, if the maximum length of stay is two years and a program has a sequence of services that requires two years for its full effect, the program will be less effective when the average length of stay is six months. Also, any program with many short-term clients that has any plans for follow-up work or data collection with former clients will quickly be overwhelmed by the growing number who need tracking; this issue is less severe if the program has mostly long-term clients.
Finally, client flow data are important when planning solutions to homelessness. If the homeless population of a community is small and stable (mostly long-term), investing in permanent supported housing will probably be a more humane and cheaper solution than maintaining people in emergency shelter. But if the same small population is largely short-term and turnover is great, finding solutions to homelessness will entail helping a group of people that in one year may be three (New York City, average length of stay=4 months), six (Philadelphia, average length of stay=2 months), or even more times the number of people who are homeless at any given time (Burt, 1994; Culhane et al., 1994).
As an evaluator by trade, the author would feel remiss if she did not point out in this paper that one critical piece of information important to planners is almost always missing, namely, information about which programs and services are effective. People always ask this question, but very few agencies are willing to spend the time and money to find out. So planners use all the information available to assess needs, and then support programs that for the most part have not been proven to have a track record of success (which does not mean they are failures, merely that we do not know which are the most effective, and cost-effective, ways to spend homeless assistance dollars). This point will ve revisited at greater length at the conclusion to this paper.
Methods for Collecting Information for Planning
Many methods exist to obtain information about the client characteristics and geographical location that planners may use to estimate need for services (for extensive descriptions of these and other methods, see Burt, 1992b). None of these is always right, or always better. Certain data needs may require specialized techniques of data collection, but it is also true that many different techniques are capable of gathering the basic types of information listed earlier in Table 2. Every data collection effort is a compromise among data needs, the expense of getting relevant information, respondents' tolerance for talking to data collectors, and the planner/researcher's abilities and resources for analysis and interpretation. This being said, Table 3 details some commonly used methods of data collection, dividing the options into those designed to obtain full counts through methods that do not rely on probability sampling, methods based on probability sampling, tracking databases, and other approaches.
| TABLE 3
COMMON METHODS FOR COLLECTING PLANNING INFORMATION |
|||
Method |
Usual Places to Find People for Study | Usual Period of Data Collection and of Estimate | Probable Complexity of Data Collected |
| Full Counts and Other Non-Probability Methods | |||
| Analysis of agency records | Specific agency | Varies; usually not done to develop a population estimate | Whatever the agency routinely records in its case documents |
| Simple count, involving significant amounts of data by observation or from minimal agency records (e.g., Boston, Nashville, Minnesota quarterly shelter survey) | Shelters, Streets | 1 night; point-in-time estimate | Enumeration, + very simple population characteristics (gender, adult/child, race) |
| Simple count with brief interview (e.g., Pasadena, Colorado) | Shelters, Meal programs, Streets | 1 night; point-in-time estimate | Enumeration + basic information as reported by respondent |
| Screener, counts and brief interviews for anyone screened in, plus unduplication using unique identifiers (Kentucky) | Service agencies of all types | Several weeks or months; point-in-time and period prevalence estimate | Enumeration + basic information as reported by respondent |
| Complete enumeration through multiple agency search and referral (Ohio, First et al.), followed by extensive interview (also unduplication) | Service agencies and key informants | Several weeks or months; point-in-time and period prevalence estimate | Usually extensive |
| Probability-Based Methods | |||
| Block probability with substantial interview (e.g., Rossi, Vernez et al., DC*MADS) | Streets | Several weeks or months; point-in-time estimate | Usually extensive |
| Other probability approaches | Abandoned buildings, conventional housing in poor neighborhoods | Several days or weeks; point-in-time estimate | Enumeration + basic information as reported by respondent |
| Service-based random sampling (e.g., Rossi; UI 1987; DC*MADS; NSHAPC) | Usually homeless assistance programs | Several weeks, months, or years; point-in-time estimate | Usually extensive |
| Shelter and other service tracking systems that allow unduplication across all services in a jurisdiction over time | Service agencies | Ongoing; point-in-time or period prevalence for periods of any length | Whatever the system collects, but usually simple data for administrative purposes |
| Other Interesting Methods | |||
| Surveys of the housed population (e.g., Link) | At home | Multi-year; produces period prevalence for periods asked about | Basic information as reported by respondent |
| Longitudinal studies (e.g., in Los Angeles, Oakland, Minneapolis, New York City) | Shelters, soup kitchens, streets | Multi-year; does not produce a population estimate | Extensive information, collected from the same person at several points in time |
| Boston-Emergency Shelter Commission, 1990; Nashville-Lee, 1989; Minnesota-Department of Children, Families and Learning, Office of Economic Opportunity; Pasadena-Colletti, 1993; Colorado-State of Colorado, 1988; Kentucky-Kentucky Housing Corporation, 1993; Ohio-First, Rife & Toomey, 1994; Rossi, Fisher & Willis, 1986; Vernez et al., 1988; DC*MADS-Bray, Dennis & Lambert, 1993; Burt & Cohen, 1989; NSHAP--Tourkin & Hubble, 1997; tracking databases-Culhane et al., 1994; Burt 1994; for ANCHoR, see www.prwt.com/anchor1; Link et al., 1994, 1995; longitudinal studies-Wong, 1997. | |||
Full-Count and Other Non-Probability Methods
Included in Table 3 as its first row is a method that is probably the most widespread of allCanalysis by each agency of its own records of service delivery, to understand past experience as a guide to the future. Some communities, as well as some whole states, have developed ways to aggregate these single-agency experiences by having each agency report common data elements to a central office that compiles the information for planning purposes. Even with such data, however, duplicate counting of individuals is often a problem at every level, and for even the shortest time periods. Soup kitchens may be able to tell you how many meals they served yesterday, but not how many people. Large shelters may be able to say how many bed-nights were used during the past week or month, but not how many people were sheltered. Even if individual agencies can produce unduplicated counts of people, once two or more agencies pool their data into community-wide or statewide reporting, there usually is no way to tell what the duplication might be across agencies either on the same day (e.g., Person A uses both a soup kitchen and a shelter on Thursday) or on different days (e.g., Person B uses Shelter 1 during week 1 and shelter 2 during week 2).
Efforts to obtain simple counts of the homeless population usually occur on a single night and within a relatively well-defined and not too large geographical area, and include searches of outdoor locations and enumerations within shelters. Early and ongoing efforts to perform such counts occurred in Boston and in Nashville (Emergency Shelter Commission, 1990; Lee, 1989). They obtained only the total number of homeless people encountered, plus some minimal descriptive data such as gender and whether the person was an adult or a child. The Office of Economic Opportunity in the Minnesota Department of Children, Families and Learning has conducted a statewide variation of this type of count four times a year since 1985, learning for a specific night each quarter how many people in each reporting area are sheltered on that night, along with whether they are men, women, or children, and whether the children are dependent or alone. The fact that these surveys occur regularly (although Nashville has stopped doing theirs) gives these jurisdictions a documented history and the ability to track trends, which can help with planning.
One variation on the simple count is to conduct a brief interview (usually about 10 minutes) with the people enumerated, taking either a random sample or everyone. This approach has been used in many places, including Pasadena, California (Colletti, 1993) and Colorado (State of Colorado, 1988). The data collection is still largely limited to one night or to one 24-hour period, and produces a point-in-time snapshot of the population. Because the interview is brief, relatively few issues related to service need can be explored in depth, but more information can be obtained with this approach than is usual with the simple count that relies heavily on observation.
The Kentucky Housing Corporation (1993) conducted another variation on the simple count. It used a brief interview as did Pasadena and Colorado, but made several methodological changes that might be of great interest to planners in other areas with relatively sparse populations and few homeless-specific services. This study greatly expanded the types of agencies through which contact was made with homeless people, including many mainstream agencies that homeless people might approach for assistance. These included health and mental health centers, jails, libraries, community action agencies, food pantries, agencies handling FEMA/EFSP funds, welfare offices, and generic social service agencies, among others. Contact with each individual approaching an agency began with a two-question screener that quickly identified the people who would need to complete the remaining 16 questions on the interview. In addition, the time frame for data collection was extended from the usual one night to two months. These changes were made to accommodate the scarcity of homeless-specific services and the different patterns of service use found in the rural areas that make up most of Kentucky. Agency contacts were supplemented by searches of outdoor locations, using homeless or formerly homeless individuals as guides. This study also had to devise a method for unduplicating the various reports of homeless people coming in from the different agencies over the two-month period, which it did by means of a unique identifier based on the last four digits of a social security number and the first four letters of the person's last name. Kentucky will repeat this survey in 1999. A more elaborate version of this methodology was pursued in rural Ohio (First, Rife & Toomey, 1994), using the same broad array of contact agencies to identify homeless people, a six-month data collection period, and a much more extensive interview.
The next set of methods described in Table 3 are those based on taking random samples and developing estimates rather than full enumerations of homeless populations. Various things can be sampled at the first stage, including city blocks or other geographic areas, abandoned buildings or conventional housing units in very low-income neighborhoods, or homeless assistance and other service programs. Once at the sampled location, individuals found there are sampled and interviewed. Block probability methods have proved to be very expensive, and are mostly not used any more since it has become clear that different versions of service-based methods will achieve as much or more coverage of the homeless population in many instances. There are exceptions, of course, for specific subpopulations among the homeless who rarely or never use services. But the way to assure coverage for these subpopulations will most likely involve visits to locations they are known to frequent rather than on random selection of blocks. The data collection sites may be randomly selected (for instance, by selecting abandoned buildings from a city’s list of tax-foreclosed properties), or they may be purposively selected as in DC*MADS' use of street “encampments” and the Course of Homelessness study’s use of known outdoor locations in downtown Los Angeles and the parking and camping areas in Los Angeles’ west side beach communities where many homeless people who did not use services could be found.
Probability-based methods take more sophistication to use than simple one-night sweeps of shelters and city streets, but their advantage is that they usually can provide more accurate estimates of the non-sheltered parts of the homeless population. And, because sampling cuts down the number of individuals one must speak with, more extensive data may be collected through interviews for the same resource commitment as would be used to try to find and/or speak with everyone. Much has been written about the advantages of these methods, so no more will be said here, except to point out that service-based random sampling could be done by local researchers for reasonable cost and could provide much useful information.
Tracking databases (usually of shelters) have received a lot of attention recently, due in part to the many articles that have appeared using the Philadelphia and New York City databases (see, e.g., Culhane et al., 1994) and to the growing interest in ANCHoR and other tracking database software. A growing number of other cities have similar types of tracking systems, including Boston, Detroit, Anchorage, Baltimore, St. Paul/Ramsey County, Minnesota, Columbus/Franklin County, Ohio, Santa Monica and San Diego, California, Ft. Worth, Texas, Denver, Colorado, the State of Rhode Island, and Maricopa County, Arizona (including Phoenix). Burt (1994) summarizes data from some of these. Interest has been growing among municipalities in developing systems that can unduplicate across programs and over time, and there has been an effort to develop “canned” systems that still contain the flexibility to be adapted to the needs of different communities. Systems developed in Denver and San Diego have been adapted by some other localities, and the U.S. Department of Housing and Urban Development has supported the development of the ANCHoR system by a group at the University of Pennsylvania and PRWT, Inc. (For information about ANCHoR, visit its website at www.prwt.com/anchor1). Approaches to actually getting the data into the community-wide system have varied, with some communities placing linked computers in every service agency, others having service agencies send hard copy to a central location for data entry, and some communities do both.
Culhane and Kuhn (1998) make a strong case for the value of this type of data for administrative purposes. They discuss the value to planners of obtaining knowledge of client flow, the distribution of short and long stays, and analysis of client characteristics among people with significantly different patterns of stay. These data would contribute to planners’ ability to estimate the potential demand for prevention and crisis services, and to informed decision making about where the system wants to put its resources.
These systems have not been without their glitches and downsides, however. They are hard to get up and running to the satisfaction of all users of the system. Some systems may be set up with an emphasis on community-wide analysis of data but individual agencies do not get much feedback that is of immediate help to them in serving clients. Other systems emphasize the control of individual agencies over their own data, which makes it valuable to each agency, but are weaker on the shared use of data and the production of systemwide statistics. Issues of privacy and data confidentiality are always present, but can be solved with concerted effort. This is important because once they start using a tracking data base system, service agencies quickly recognize the value to their clients of being able to share information about the client with other agencies. But if the system has been set up with maximum privacy protections, this sharing may be difficult to achieve in retrospect. Another disadvantage of current systems is that few include any services other than shelters. Maricopa County, Arizona is an exception, as it includes a large health care for the homeless site that serves many street homeless people and also asks about the homed or homeless status of people using other agencies in the system such as Head Start and community action agency programs.
A final issue that is beginning to arise in some communities with several years of experience with working data bases has to do with getting paid. Agencies have come to realize that some of their budget comes in the form of reimbursement for services given to clients, and that these services are being registered in the data system. There are anecdotes that agencies have become possessive of their clients, possibly up to the point of not referring them to other agencies for services because they want all the "credit" for that client. Communities installing tracking data bases and intending to use them as one element in funding decisions would do well to address these issues of ownership and sharing directly, so that clients get the appropriate services from the appropriate providers.
The last two data collection methods included in Table 4 are unlikely to be conducted by local or state planners, but the information from the original studies should be of great interest. These include national (or more limited) telephone surveys of households using random digit dialing, and longitudinal studies of homeless populations. The first method can be used to get estimates of lifetime and recent homeless experiences of currently housed people, while the second shows us the patterns of entering and leaving homelessness over extended periods of time.
The estimates produced by shelter tracking databases of the proportion of whole city populations that have experienced homelessness, hovering around 3 percent of the population over a three-year period, are supported by results from a completely different source-household telephone surveys using random digit dialing conducted by Link and his team. Their results are that about 3 percent of American adults (7 to 8 million people) experienced literal homelessness within the past five years (Link, Susser et al., 1994; Link, Phelan et al., 1994).
Obviously, most of the homeless episodes tallied by the shelter tracking and the household survey data did not last a very long time, or the one-day homeless population would be much higher than the 500,000 to 600,000 commonly thought to be a reasonable estimate for a 24-hour period. The new results have made both researchers and policy makers think again about what might be the best approach to serving homeless people, and to consider what services might be relevant for someone who just needs a little help to leave homelessness or for whom appropriate interventions might prevent homelessness, as well as for someone who needs a lot of help.
Longitudinal studies of homeless cohorts became available in the 1990s for the first time. Several research projects (in Minneapolis, Minnesota, Los Angeles and Oakland, California, and New York City) followed a sample of homeless people over extended periods of time. These efforts (see Koegel & Burnam, 1991; Koegel, Burnam & Morton, 1996; Piliavin, Sosin & Westerfelt, 1993; Robertson, Zlotnick & Westerfelt, 1997, Schinn, 1997) reveal in great detail the complexity of homeless careers. While some people may have only one homeless episode, during which they are “on the streets” for the entire time, many people who are homeless at the time a sample is taken have moved in and out of housing frequently, depending on their available funds and other supports.
The results of longitudinal research studies help us understand many things about homeless careers. On one hand, they help us to see how many people experience single short spells of homelessness and are able to leave on their own and never return. These people may never draw much attention from service providers and planners because they do not draw heavily on service resources. Nevertheless, their experiences can help us understand the circumstances that allow people to leave homelessness and stay housed, and may also be important when planning prevention efforts.
On the other hand, these longitudinal studies help us to see the difficulties encountered by another set of people who find it very hard to leave homelessness for good, and what it will take to truly end this type of homeless career. Longitudinal studies have documented some of the near-term causes of homeless episodes, and shown just how fragile is the hold some people have on stable housing. Planners should be aware of these results as they think through what continuum of services they want to create in their communities.
Without knowing what works, planners with the best information in the world about the service needs of homeless people will not be able to make the best decisions about which programs are the best investment of local resources. Information about program performance and impact is relatively scarce in the homeless services arena (which does not make homeless assistance services any different from most other service arenas). Further, the information that we do have is skewed to particular types of programs for particular segments of the homeless population. For the most part, we have the best information about programs for people with mental illness and substance abuse problems and minimal information about the effectiveness of services for anyone else, including families. Other papers in this symposium go into much greater depth on issues of service effectiveness than there is space for here, but some information about “what works” is essential here because it is so critical for decision makers to know, and so rarely available, that it would be inappropriate for anyone to think they had all necessary information just because they were able to describe their homeless population.
We know a good deal about how to serve homeless people with mental illness, drug abuse, or alcoholism because several provisions of the Stewart B. McKinney Homeless Assistance Act of 1987 directed federal government agencies to sponsor relevant service research projects. Portions of the Act authorized funding to identify effective models of care that could maintain these most difficult-to-help long-term homeless people in stable housing situations. The evaluation research was funded through the National Institute of Mental Health and the National Institute on Alcohol Abuse and Alcoholism (Fosburg et al., 1996; Morrissey et al., 1996; National Resource Center on Homelessness and Mental Illness, 1992; Randolph et al., 1996; Shern et al., 1997; Sosin et al., 1994; Tessler and Dennis, 1989).
The first, most remarkable thing we know is that the programs do work. Many of them have been able to retain around 80 percent of the previously homeless people they serve in decent, stable housing arrangements. We also know that without services attached, they do not work. The critical services needed are: negotiating with landlords and neighbors, handling situations of decompensation or slipping off the wagon, assuring that the rent is paid and the housing is kept clean, and supplying tangible goods when necessary such as furniture, transportation, and food. These critical services are not readily available from other agencies in the community, nor are they the responsibility of any other agency. Therefore, they tend to be absent if federal funds do not cover them. Local decision makers would do well to consider supporting these services with local funds if they want to create maximally effective residential programs for their hardest-to-serve chronically homeless population.
Further, we know that without services, not only do the previously homeless people with serious disabilities lose their own current housing, but they lose it in a way-by antagonizing landlords and neighbors-that the housing unit itself is likely to stay lost and unavailable for other homeless persons. Thus the program wastes the energy and resources already invested in finding and arranging the housing, and has to start over with a bad track record. This is wasteful for all concerned, and does little to build community good will toward homeless people with severe disabilities. Local planners may want to assess the wisdom of spending funds for housing but not including the supportive services that make housing investments successful.
The limited amount of research available on service outcomes for homeless families (Rog and Gutman, 1997; Shinn, 1997; Wong, Culhane and Kuhn, 1997) indicates the efficacy of providing housing subsidies as a means of stabilizing residential patterns among homeless families and suggests that without such subsidies, these families’ personal resources, skills and human capital are not adequate to maintain themselves in housing and otherwise take care of family responsibilities. These are also the families likely to be the least capable of finding employment at the level of self-sufficiency, and therefore to be the hardest hit by welfare reform provisions limiting the time of welfare receipt. Loss of welfare income may precipitate episodes of homelessness.
We have learned a great deal about homeless populations in the past decade and a half, and have learned even more about how to learn about them. Many of the methods described in this paper can be adapted for use at the local and state levels, where they could produce extremely valuable information for planning purposes. Most of the methods, once beyond simple counts, can supply decision makers with a great deal of data about the characteristics of homeless people using services in a community. These characteristics extend far beyond the simple demographics described above, and include the presence of various disabling conditions that can be used to design the specialized services most appropriate to the local population.
More and more communities are coming to recognize the value of good data for rationalizing their service programs for homeless people. When you go to a community that has installed a tracking data base, for example, they are most likely to tell you that the data don’t resolve all of their priorities or make all of their decisions. But since they have had the data, they say, they no longer spend any time arguing about the scope of the problem (which they used to do all the time), and can focus their efforts on deciding what to do about it.
The newest types of data, in particular the tracking data bases, have raised many important policy issues that were semi-invisible before. We now know, or could know, the proportion of homeless spells that are very short term and the characteristics of the people who have them. This information could help us design appropriate emergency services, including some that would not require a person to become literally homeless (i.e., to enter a shelter) just to access them. By the same token, we now know, or could know, the proportion of homeless spells that are very long-term, the characteristics of the people who have them, and the amount of system resources they absorb. This information could help us to decide that there are better, and even cheaper, ways to help these people through stable, supported permanent housing arrangements.
The episodic group among the homeless is the most interesting, because its picture is least developed. Culhane, because his data source is shelter stays, calls people episodic when they go in and out of shelter regularly. But perhaps they are not episodic in the sense that they go in and out of homelessness; they could merely move to the streets and back again to shelter. Other types of data would be more capable of exploring different patterns of episodic homelessness.
In addition, we should ask what we mean by "episodic," as the word could have a number of different meanings. Longitudinal studies help us to understand what some of these meanings might be. People whose incomes last them only three weeks out of every month could be in hotels or motels for those three weeks, and in shelters or on the streets for the rest of the month. This is a pattern that combines both an episodic element and a long-term element (they have been doing this for years). In shelter tracking data bases using a 30-day exit criterion, all of these people would be counted a continuous stayers (they would never be out of shelter for a period greater than 30 days), but this pattern may not be what we intuitively mean when we speak of “long-term chronic”. Knowledge of patterns of service use may stimulate a community to ask itself what it is really trying to accomplish with its services, and perhaps to design better ways to intervene in pursuit of those goals.
Finally, it bears mentioning that we are living in a time when major streams of income support for very poor individuals and families are being eliminated outright (General Assistance at the state and local levels) or limited and restricted to certain people, for certain time periods, and contingent upon certain prescribed behaviors (Temporary Assistance to Needy Families, formerly Aid to Families with Dependent Children; Food Stamps). Anecdotes about how well welfare reform is “working” are balanced by anecdotes about individuals and families who have lost benefits and become homeless. It will be important in the coming years to document the effects of the fraying safety net on the abilities of people to remain housed or to leave homelessness once in that condition.
Barrett, D. F., Anolik, I. & Abramson, F. H. (1992). The 1990 Census Shelter and Street Night Enumeration. Bureau of the Census, Decennial Statistical Studies Division. Paper presented at the American Statistical Association Annual Meetings, Boston, MA: August 10.
Bassuk, E. L., Buckner, J. C., Weinreb, L. F., Browne, A., Bassuk, S. S., Dawson, R. & Perloff, J. N. (1997). Homelessness in Female-Headed Families: Childhood and Adult Risk and Protective Factors. American Journal of Public Health, 87(2): 241-248.
Bray, R. M., Dennis, M. L. & Lambert, E. Y. (1993). Prevalence of Drug Use in the Washington, DC Metropolitan Area Homeless and Transient Population: 1991. Rockville, MD: National Institute on Drug Abuse, Division of Epidemiology and Prevention Research.
Burt, M. R. (1992a). Over the Edge: The Growth of Homelessness in the 1980s. New York, Russell Sage Foundation.
Burt, M. R. (1992b). Practical Methods for Counting Homeless People: A Manual for State and Local Jurisdictions. Washington, D.C.: Interagency Council on the Homeless and the Department of Housing and Urban Development.
Burt, M. R. (1994). Comment on Dennis P. Culhane, Edmund F. Dejowski, Julie Ibañez, Elizabeth Needham, and Irene Macchia's "Public Shelter Admission Rates in Philadelphia and New York City: The Implications of Turnover for Sheltered Population Counts." Housing Policy Debate, 5(2): 141-152.
Burt, M. R. & Cohen, B. E. (1989). America's Homeless: Numbers, Characteristics, and the Programs that Serve Them. Washington, DC: Urban Institute Press.
Burt, M. R., L. Y. Aron, D. D. Chaplin, and S. E. Cavanagh. (1993). The Homeless in the 1990 Decennial Census: Description of Persons Found in Emergency, Runaway Youth, and Battered Women's Shelters, Visible in the Streets, and in Other Selected Group Quarters. Washington, DC: Urban Institute.
Caton, C. L. M., Shrout, P. E., Eagle, P. F., Opler, L. A., Felix, A., & Dominguez, B. (1994). Risk Factors for Homelessness Among Schizophrenic Men: A Case-control Study. American Journal of Public Health, 84: 265-270.
Colletti, Joseph. (1993). The City of Pasadena Homeless Count: September 23, 1992. San Gabriel Valley, CA: Lutheran Social Services of Southern California.
Culhane D. & Kuhn R. (1998). Patterns and Determinants of Public Shelter Utilization Among Homeless Adults in New York City and Philadelphia. Journal of Policy Analysis and Management, 17(1): 23-43.
Culhane, D. P., Dejowski, E. F., Ibanez, J., Needham, E., & Macchia, I. (1994). Public Shelter Admission Rates in Philadelphia and New York City: The Implications of Turnover for Sheltered Population Counts. Housing Policy Debate, 5: 107-140.
Culhane, D. P., Lee, C. M., & Wachter, S. M. (1996). Where the Homeless Come From: A Study of the Prior Address Distribution of Families Admitted to Public Shelters in New York City and Philadelphia. Housing Policy Debate, 7(2): 327-365.
Emergency Shelter Commission. (1990). State of Homelessness in the City of Boston, Winter 1989-1990. Boston: Author.
First, R. J., Rife, J. C., & Toomey, B. G. (1994). Homelessness in Rural Areas: Causes, Patterns, and Trends. Social Work, 39(1): 97-108.
Fosburg, L. B., Locke, G. P., Peck, L. & Finkel, M. (1997). National Evaluation of the Shelter Plus Care Program. Washington, DC: U. S. Department of Housing and Urban Development, October.
Herman, D. B., Susser, E. S., Struening, E. L. & Link, B. L. (1997). Adverse Childhood Experiences: Are They Risk Factors for Adult Homelessness? American Journal of Public Health, 87(2): 249-255.
Kentucky Housing Corporation. (1993). Kentucky Homeless Survey Preliminary Findings. Lexington, KY: Kentucky Housing Corporation.
Koegel, P. & Burnam, M. A. (1991). The Course of Homelessness Study: Aims and Designs. Paper presented at the 119th Annual Meeting of the American Public Health Association, November 1991, Atlanta, GA. Santa Monica, CA: RAND Corporation.
Koegel, P., Burnam, A., & Morton, S. C. (1996). Enumerating Homeless People: Alternative Strategies and Their Consequences. Evaluation Review, 20: 378-403.
Koegel, P., Melamid, & Burnam, M. A. (1995). Childhood Risk Factors for Homelessness Among Homeless Adults. American Journal of Public Health, 85(12): 1642-1649.
Kuhn, R. & Culhane, D. P. (1998). Applying Cluster Analysis to Test a Typology of Homelessness By Pattern of Shelter Utilization: Results From the Analysis of Administrative Data. American Journal of Community Psychology, 26(2): 207-232.
Lee, B. (1989). Stability and Change in an Urban Homeless Population. Demography, 26(2): 323-334.
Link, B. G., Susser, E., Stueve, A., Phelan, J., Moore, R. E., & Struening, E. (1994). Lifetime and Five-Year Prevalence of Homelessness in the United States. American Journal of Public Health, 84: 1907-1912.
Link, B., Phelan, J., Bresnahan, M., Stueve, A., Moore, R. & Susser, E. (1995). Lifetime and Five-Year Prevalence of Homelessness in the United States: New Evidence on an Old Debate. American Journal of Orthopsychiatry, 65(3): 347-354.
Mangine, S. J., Royse, D. & Wiehe, V. R. (1990). Homelessness Among Adults Raised As Foster Children: A Survey of Drop-in Center Users. Psychological Reports, 67: 739-745.
Morrissey, J., Calloway, M., Johnsen, M. & Ullman, M. (1997). Service System Performance and Integration: A Baseline Profile of the ACCESS Demonstration Sites. Psychiatric Services, 48(3): 374-380.
National Resource Center on Homelessness and Mental Illness. (1992). Annotated Bibliography: Developing Housing for Homeless Persons with Severe Mental Illnesses. Delmar, NY: Policy Research Associates, Inc., March and September. (1 800 444-7415).
Piliavin, I., Sosin, M. & Westerfelt, A.H. (1993). The Duration of Homeless Careers: An Exploratory Study. Social Service Review, 67: 576-598.
Randolph, F., Blasinsky, M., Leginski, W., Parker, L. B., Goldman, H. H. (1997). Creating Integrated Service Systems for Homeless Persons with Mental Illness: The ACCESS Program. Psychiatric Services 48(3): 369-373.
Ringwalt, C. L., Greene, J. M., Robertson, M. & McPheeters, M. (1998). The Prevalence of Homelessness Among Adolescents in the United States. American Journal of Public Health, 88(9): 1325-1329.
Robertson, M. J., Zlotnick, C., & Westerfelt, A. (1997). Drug Use Disorders and Treatment Contact Among Homeless Adults in Alameda County, California. American Journal of Public Health, 87 (2): 217-220.
Rog, D. J. & Gutman, M. (1997). The Homeless Families Program: A Summary of Key Findings. In S. Isaacs & J. R. Knickman (Eds.), To Improve Health and Health Care, 1997: The Robert Wood Johnson Anthology. San Francisco: Jossey-Bass Publishers, 209-231.
Rossi, P. H. (1994). Troubling Families: Family Homelessness in America. American Behavioral Scientist, 37: 342-395.
Rossi, P. H., Fisher, G. A. & Willis, G. (1986). The Condition of the Homeless in Chicago. Amherst, MA and Chicago: Social and Demographic Research Institute and the National Opinion Research Center.
Shern, D. L., Felton, C. J., Hough, R. L., Lehman, A. F., Goldfinger, S., Valencia, E., Dennis, D., Straw, R., & Wood, P. A. (1997). Housing Outcomes for Homeless Adults with Mental Illness: Results from the Second-Round Mckinney Program. Psychiatric Services, 48: 239-241.
Shinn, M. (1997). Family Homelessness: State or Trait? American Journal of Community Psychology, 25(6): 755-769.
Sosin, M. R., Yamaguchi, J., Bruni, M., Grossman, S., Leonelli, B., Reidy, M. & Schwingen, J. 1994. Homelessness and Substance Abuse in Community Context: A Case Management and Supported Housing Demonstration. Chicago: University of Chicago, School of Social Services Administration.
State of Colorado. (1988). Report of the Governor’s Task Force on the Homeless. Denver, CO.
Susser, E. S., Lin, S. P., Conover, S. A. & Struening, E. L. (1991). Childhood Antecedents of Homelessness in Psychiatric Patients. American Journal of Psychiatry, 148: 1026-1030.
Susser, E. S., Struening, E. L. & Conover, S. A. (1987). Childhood Experiences of Homeless Men. American Journal of Psychiatry, 144: 1599-1601.
Tessler, R. C. & Dennis, D. L. (1989). A Synthesis of NIMH-funded Research Concerning Persons Who Are Homeless and Mentally Ill. Rockville, MD: National Institute of Mental Health, Division of Education and Service Systems Liaison, Program for the Homeless Mentally Ill.
Tourkin, S., & Hubble, D. (1997). National Survey of Homeless Assistance Providers and Clients: Data Collection Methods. Suitland, MD: U. S. Department of Commerce, Bureau of the Census.
Vernez, G., Burnam, M. A., McGlynn, E. A., Trude, S. & Mittman, B. (1988). Review of California's Program for the Homeless Mentally Disabled. Santa Monica, CA: The RAND Corporation.
Weitzman, B. C., Knickman, J. R. & Shinn, M. (1992). Predictors of Shelter Use Among Low-Income Families: Psychiatric History, Substance Abuse, and Victimization. American Journal of Public Health, 82: 1547-1550.
Wong, Y.-L. I., Culhane, D. P., & Kuhn, R. S. (1997). Predictors of Exit and Reentry Among Family Shelter Users in New York City. Social Service Review, 71: 441-462.
Wong, Y-L. I. (1997). Patterns of Homelessness: A Review of Longitudinal Studies. Understanding Homelessness: New Policy and Research Perspectives. Washington, DC: Fannie Mae Foundation.
Wood, D. L., Valdez, R. B., Hayashi, T., & Shen, A. (1990). Homeless and Housed Families in Los Angeles: A Study Comparing Demographic, Economic, and Family Function Characteristics. American Journal of Public Health, 80(9): 1049-1052.
(1) The number of homeless people in Louisville is probably underrepresented by these data, because Louisville reported only its sheltered population at a point in time, it did not use the overall study method of a two-month data collection period and a variety of agencies, nor were there searches of outdoor locations in Louisville.
Return to Table of Contents, ASPE home page, HHS home page.