by
Marybeth Shinn, Ph.D.
Jim Baumohl, D.S.W.
The General Accounting Office's 1990 conclusion about the prevention of homelessness still holds: It remains "too early to tell" what works best. Eviction prevention programs show some promise but have not been rigorously evaluated and tend to exclude people at highest risk of homelessness. Several studies suggest that individuals with severe mental illness can be supported in the community, but the mixture of housing and supportive services necessary remains unclear. There is even less evidence for the usefulness of planning discharges from institutions or of programs to ameliorate domestic conflicts. However, even if expanded to reach 100 percent of their target populations and even if 100 percent successful, all of these programs together would reach only a minority of the people who become homeless each year, and targeting efforts would yield many false alarms for each future case of homelessness correctly identified. Based on evidence that subsidized housing, with or without supportive services, is sufficient to end homelessness for most families, and given the important role of subsidized housing (everywhere it has been examined) in ending homelessness among people with serious mental illnesses, we propose a shift to selected strategies of prevention, such as providing housing subsidies to those with worst—case housing needs, supporting employment and transitional assistance to poor, young people setting up households for the first time, and focusing efforts on communities from which large proportions of homeless people originate.
Anyone who has passed a person sleeping in a doorway, seen a family with their belongings heaped in a shopping cart, observed makeshift dwellings under a bridge, or visited a shelter where strangers lie warily on adjacent beds, is likely to have thought that, surely, such scenes could be prevented. In our view, homelessness in the United States could be avoided, for the most part; and yet we are not sanguine about the prospects. A lack of resources, even in the midst of the century's most sustained peace-time economic growth, is not the only obstacle, though it is the most formidable. In addition, many current efforts to prevent homelessness may not be based on sound premises. Moreover, tributes to their effectiveness are statements of faith that cannot withstand serious, scientific scrutiny.
In this essay we cast a critical eye on existing measures to prevent homelessness. By way of further introduction, we discuss the logic and basic terminology of prevention. Next, paying careful attention to the conceptual and methodological issues involved, we review research on programs that aim to prevent homelessness. We conclude that most such efforts do useful things for needy people, but they seem to have only marginal impact on the prevention of homelessness. In view of the conceptual and methodological quandaries we identify, and the empirical findings we review, we recommend that homelessness prevention be re-oriented from efforts to work with identified at-risk persons to projects aimed at increasing the supply of affordable housing, sustainable sources of livelihood, and the social capital of impoverished communities.
Simply put, to prevent means to keep something which would have happened from happening in fact. At a minimum, the logic of prevention requires that its critical terms be specified in detail and that we be able to tell if the effort has been successful. Thus, we must be able to define clearly what is to be prevented, we must be able to specify the intervention(s), and we must be able to establish a causal (or at least correlational) connection between intervention and the demonstrated avoidance of the undesirable phenomenon. Other things equal, the more narrowly we define what is to be prevented, the more elegant the intervention, and the more rigorous the experimental design of evaluation, the easier the task of determining effectiveness. Thus, the prototypical example is of a discrete disease entity (say, polio), preventable by vaccination (a simple, easily standardized intervention), where effectiveness can be demonstrated clearly by comparing outcomes in vaccinated and unvaccinated samples.
Alas, most unwanted phenomena are more like suicide than polio. They have ambiguous definitions (when is a suicide attempt "real" and thus to be counted as a case?); they have multiple causes; they are only somewhat responsive to a variety of interventions; and outcomes are difficult to assess by rigorous means (the determination of suicide after the fact is notoriously problematic). Moreover, most interventions are complex, difficult to standardize, and often implicated in an unwitting redefinition of the phenomenon by virtue of the rewards or penalties they distribute. (For example, a "right to shelter" provision may cause some people living in crowded or deficient housing to present themselves at shelters, thus redefining their circumstances and the nature of what we call homelessness.) Only in the most strained metaphors are social interventions anything like vaccinations.
Note, too, that prevention involves predicting the future. To determine whether an intervention is successful, we must know the likelihood that the unwanted will occur so that we may compare this with the actual outcome following intervention. Not everyone will get even an easily transmissible disease. In the case of a relatively rare phenomenon, few will be affected. Thus, to allocate resources efficiently, or to ration scarce resources, prevention programs target particular subjects who have been "exposed" (in the language of disease) or who are, by some theoretically plausible or empirically determined criteria, "at risk" of being affected. An example of effective targeting is the prevention of mental retardation due to phenylketonuria. A simple, inexpensive blood test, shortly after birth, accurately identifies infants who lack the enzyme that metabolizes certain proteins. Children can be treated successfully with special diets low in the amino acids that give rise to phenyls.
Unfortunately, most unwanted phenomena are not much like phenylketonuria. There is no one factor that accurately predicts them. Rather, there are usually a number of predictive correlates, with risk increasing as the number of such risk factors (the overall burden of risk) increases. Even so, the accuracy of such prediction often is not particularly high. This results in poor targeting and consequent inefficiencies in prevention programs even when the interventions work as intended. But the example of phenylketonuria is instructive in one important respect: Although the problem is the lack of an enzyme, the effective prevention measure does not replace the enzyme but modifies the child's diet instead. We will suggest that in the case of homelessness as well, the solution may not always match the problem directly.(1)
Prevention programs are of three ideal types (Mrazek & Haggerty, 1994, following Gordon, 1983). Universal prevention programs are available to the entire population, although they are sometimes targeted at people who have reached a particular period of life. Such programs may be narrow and inexpensive, such as childhood immunizations to prevent measles, or quite expensive and expansive, such as old-age pensions, intended to prevent poverty among the elderly; subsidized housing programs intended to prevent homelessness; or the wholesale construction of water treatment facilities to prevent water-borne disease. As these examples suggest, prevention programs (of all types) may involve strengthening individuals (a measles vaccine) or changing the environment (water treatment).
Selected prevention programs are aimed at people at risk due to membership in some group. No individual screening is required for participation. For example, an educational program might be aimed at occupational groups at risk of repetitive motion injuries.
Indicated prevention programs are directed to people at risk because of some individual characteristic or constellation of characteristics. Individual-level screening is required. Programs to mitigate the consequences of genetic diseases are of this sort.
Ideal types are heuristic devices, of course, and distinctions among types of prevention are often fuzzy. For instance, people discharged from mental hospitals comprise a group at risk of homelessness, but also have the individual risk factor of prior mental hospitalization. A universal housing program may, in fact, be attractive only to those who are poor (an individual risk factor).
Selected and indicated strategies may be more efficient than universal measures when it is easy to identify and deliver interventions to groups of people or individuals at risk for a particular condition. The efficacy of targeting is thus of fundamental importance to the design of prevention programs, and the costs of targeting must be compared with the costs of offering programs more broadly or allowing people to select themselves for universal programs attractive only to those with high levels of perceived need.
A prevention program is at least somewhat effective if it reduces the overall incidence of a problematic condition (the number of people who newly become affected over some defined period) or its prevalence (the number of people affected at a particular point in time, or over some defined period). Showing that most people who use the program do not become affected is insufficient. Perhaps they would not have been affected in any case. Perhaps the condition has been delayed, but not averted. Or perhaps some aspects of the intervention have encouraged or allowed consumers of the program to simply "jump the queue" to receive services, so that others, pushed back in the queue, are at greater risk.
Finally, programs can focus on preventing new cases of something (usually called primary prevention) or on the early identification and treatment of current cases (usually called secondary prevention). Secondary prevention efforts may reduce the prevalence of a condition, but they do not reduce its incidence.
Later in this essay we will consider whether homelessness prevention programs fulfill these logical requirements. Here, though, we need to establish the operational definition of our subject. If homelessness is the undesirable phenomenon we wish to prevent, we need to specify it, for without a definition, we will not be able to identify the subjects ("targets") of our efforts, nor will we be able to identify the presence or absence of the condition.
The simplest approach is merely to adopt the conventions of the federal government and most survey researchers (see Burt, 1996). Indeed, since the programs we will review are guided by these conventions, this is a sensible restriction for the purposes of this paper. Simply stated, people are homeless when they live without housing or take up residence in shelters. People are "at risk" of homelessness when they have lost security of tenure in any residential setting, whether a household or an institution. Typically, homelessness prevention programs are concerned with preventing shelter entry, a criterion that is amenable to relatively easy measurement and encompasses a major public cost of homelessness even if it fails to capture private burden.
Still, consider some of the important questions begged by this definition with respect to what constitutes prevention. The size of the shelter population is driven in large part by the number of beds available and policies regarding access to them; that is, admission criteria, limits on length of stay, and so forth (e.g. Culhane, Lee & Wachter, 1996). If a shelter turns applicants away, or evicts residents after some fixed period, the count of homeless people may be limited, but it is not clear that those denied access are better off even if they do not end up on the street. Indeed, subjectively they are not, for they preferred shelter to whatever other arrangements were available. Should we then say that homelessness has been prevented for those who are denied shelter but find some arrangement short of literal homelessness? Similarly, if officials intentionally make entry into a shelter system aversive, so that some people who would otherwise apply for shelter decide to stay in overcrowded or deficient housing, should we say that homelessness has been prevented?(2)
Further, studies that follow the "careers" of homeless people over time show that, for single adults, "the state of homelessness appears to be more a drift between atypical living situations and the street than between normality and street life" (Sosin, Piliavin & Westerfelt, 1990:171). The atypical situations, such as staying with friends, do not appear to be sustainable. Should we say that homelessness has been prevented if people make the rounds of friends and family, constantly doubling up in precarious situations? Similarly, many homeless people with severe mental illnesses and substance abuse problems travel "institutional circuits" that include mental hospitals, prisons, or jails as well as shelters, shared or doubled-up arrangements, and the street (Baumohl, 1989; Hopper, Jost, Hay, Welber & Haugland, 1997; Milofsky, Butto, Gross & Baumohl, 1993; Snow & Anderson, 1993; Spradley, 1970; Wiseman, 1970). Has homelessness been prevented if people are temporarily moved from one such housing status to another?
Thus, criteria for successful prevention are established within the boundaries of the problem's definition. The operational definition of homelessness employed here is more convenient than theoretically satisfactory, but it will serve. (For a theoretical reformulation, see Hopper & Baumohl, 1994, 1996.)
Many studies have identified factors that reliably distinguish people who are homeless from some comparison group. Lindblom (1991: 963) suggested targeting prevention assistance based on a profile of risk factors that could "identify the lion's share of those extremely poor persons who will enter or reenter homelessness if they do not receive outside help." He also noted (1996: 188) that "only a small portion of the persons targeted by many prevention programs are actually at risk of repeated or prolonged homelessness." We know of only one study of the efficacy of targeting or forecasting the onset of homelessness (Knickman & Weitzman, 1989; Shinn, Weitzman, Stojanovic, Knickman, Jimenez, Duchon, James & Krantz 1998). This study examined 20 potential factors, including measures of demographic characteristics, persistent poverty, behavioral disorders, social ties, and housing that might distinguish between families on welfare who requested shelter in New York City from other New York City families in the public assistance caseload. Families who had used shelter previously were excluded from both groups. The authors used these factors to construct various multi-predictor models to forecast homelessness. Although 18 factors were related to homelessness, taken one at a time, the "best" multivariate model included 10 predictors. These variables made reliable contributions to the prediction of homelessness in the context of the other variables in the model.(3)
The model yielded a score summarizing risk for each family in the study. Choosing who should be eligible for a prevention program corresponds to choosing some cutoff for risk scores: Families with scores higher than the cutoff would be eligible; families with scores lower than the cutoff would not be served. A liberal cutoff score, selected to deliver prevention services to a large portion of those who would otherwise become homeless, also targets many families who would not become homeless in the absence of services ("false alarms"). A conservative cutoff yields fewer false alarms but also has a lower "hit rate;" that is, it reaches fewer of those who would become homeless without preventive efforts. Thus, a plot of hit rates versus false alarm rates for different predictive models is a very useful policy tool (Camasso & Jagannathan, 1995; Swets, 1973 & 1988). Shinn et al. (1998) found that the best model was able to correctly "hit" 66 percent of welfare families who requested shelter with a false alarm rate of only 10 percent.(4)
While this ratio of hits to false alarms may sound good, the population to which the false alarm rate refers is far larger than the group who will end up in shelter. At the time the data were collected, there were about 290,000 families on welfare (over the course of a year) in New York City, and about 90 percent of the approximately 10,000 families who first entered shelter over the course of the year came from the welfare caseload. Thus, to correctly reach 6,000 families (90% of 66% of 10,000), a prevention program would have to offer services to 27,000 families (10% of 290,000, less those with previous shelter experience) who would not become homeless. With respect to preventing shelter entry, over 80 percent of the services would be wasted, although such help might be valuable to families for other reasons. A more narrowly targeted prevention program that confined false alarms to 2 percent of the public assistance caseload and reached only 36 percent of those applying for shelter, would still "waste" three-fifths of its services (correctly identifying 3,600 families against 5,400 false alarms). To reach three-quarters of families applying for shelter in New York City alone, one would need to target an additional 60,000 families who would not become homeless. In addition to the problem of wasting services on those who will not become homeless, there is the problem of failing to serve those who will become homeless. Even a targeting cutoff that wastes 80 percent of the services misses 34 percent of the families who in fact become homeless.
In addition, the best predictive model included some risk factors, such as childhood disruptions or domestic violence in adulthood that might be hard to assess or verify. If access to an attractive prevention program (such as subsidized housing or valued social services) depended on such risk factors, and the prediction formula became even roughly known (as it inevitably would), the targeting effort would create incentives for people to dissemble in order to obtain services and could create an adversarial relationship between service providers, charged with certifying eligibility, and their clients. Likely, reports of the key risk factors would increase, more people would be deemed eligible for services, and the predictive power of the model would decline.
Interestingly, a model with only seven easily verified predictors did almost as well as the full model at intermediate levels of risk (65% versus 66% hits at 10% false alarms, among families on public assistance). The model included the same two demographic characteristics as the best model and all five housing variables. However, this model did less well for narrow targeting and includes one factor (race) on which it would be illegal to base access to services.
Prevention programs could be developed based on predictive models targeting other groups (such as single adults or adolescents homeless on their own), and the New York model might not apply to families in other areas (correlates of homelessness vary by locale). Still, without continuously renewed data and analysis, all such predictive models are static. This is important, because unlike the case of phenylketonuria, the correlates of homelessness shift over time, both because the phenomenon itself changes (homelessness today is not exactly like the mass dispossession of the Great Depression or the more ambiguous homelessness of post-war skid rows), and because routes to shelter residence change, thus reconfiguring the populations found there (Hopper & Baumohl, 1994, 1996). Any predictive model, then, is in jeopardy of becoming rapidly outdated and progressively inefficient. Today, most of what we know about correlates of homelessness comes from studies conducted nearly a decade ago, when economic conditions, for instance, were very different; and today's knowledge may not apply in the future, when, for example, a smaller fraction of the poor is eligible for welfare support. Homelessness is a dynamic phenomenon, chased but never really captured by research.(5)
One general lesson can be learned from the New York experience: A prevention program aimed at people with any single characteristic, such as those being discharged from mental hospitals, is likely to target only a small portion of all who become homeless; even sophisticated multivariate models with very narrow targeting (which therefore reach a very small proportion of those who become homeless) are likely to have far more false alarms than hits. A second lesson, perhaps less general, is that in the case of New York families, targeting based primarily on their housing status did about as well as models that took into account less verifiable indicators of individual risk.
If the outcome criterion to be predicted were months in shelter (which is more closely associated with public dollars spent than is simple shelter entry), it might be possible to develop more efficient predictive models. For example, Culhane and Kuhn (1998) show that in New York, 18 percent of single adult, first-time shelter users consumed 53 percent of the total days in shelter for first-time users in their first year. In Philadelphia, 10 percent consumed 35 percent of these days. The authors describe several individual factors associated with longer stays and repeat use of shelter (age, mental health and substance abuse problems, and sometimes medical conditions), but do not tell us how efficiently these high-consumers can be identified. (Such knowledge is crucial to the practical application of such data, of course.) Further, the criterion of months in shelter is problematic for another reason: The timing of exits from shelter depends in part on resources made available to residents. For example, in Philadelphia, people with serious mental disorders exited more quickly, whereas those with less serious disorders exited more slowly than those with no assessed disorder, probably because those with the most serious disorders were eligible for specialized services (Culhane & Kuhn, 1998). For families in New York, months in shelter was positively associated with subsequent stability in housing because a long time in shelter reflected movement to the top of the queue for subsidized housing (Shinn et al., 1998).
Most actual prevention programs use simple targeting strategies. A majority of the over 400 prevention programs receiving funds from the Emergency Shelter Grants Program in fiscal year 1991 used receipt of an eviction notice (52%) and/or a utilities shut-off notice (27%) to identify clients eligible for prevention services. (A program could use more than one criterion.) Sixteen per cent targeted victims of domestic violence (Feins, Fosburg & Locke, 1994a:116).
These programs are likely to reach only a small proportion of people who would otherwise become homeless. For example, among New York families requesting shelter, only 22 percent of first-time users had ever been evicted from any apartment (and many of these were informally evicted by the people with whom they were staying rather than receiving formal eviction notices from landlords), whereas 44 percent had never even had an apartment of their own for as long as a year since having children (Weitzman, Knickman & Shinn, 1990). By way of comparison, 6 percent of the public assistance caseload who had never used shelter also reported having been evicted. Because the public assistance caseload is far larger than the group of families who entered shelter, we estimate that a program that targeted welfare families facing eviction in New York would serve four or more families who would avoid entering shelter anyway for every family who would in fact enter shelter in the absence of the program, while reaching only one-fifth of the shelter population.(6)
The proportion of homeless families who have been evicted varies by time and location. Wood, Valdez, Hayashi and Shen (1990) found 34 percent with housing problems, including eviction, in Los Angeles (and many more with economic problems); Bassuk, Buckner, Weinreb, Brown, Bassuk, Dawson, and Perloff (1997) found that 26 percent of homeless families in Worcester—and 17 percent of housed poor families—had been evicted or locked out, suggesting that eviction prevention services would be far less efficient than in New York. Other studies, reviewed by Bueno, Parton, Ramirez, and Viederman (1989:8-9) reported percentages of homeless families who had been evicted ranging from 14 percent to 57 percent, with the high figures sometimes including other housing problems such as non-payment of rent. These studies did not give proportions of people evicted who ended up in shelter. Among residents of shelters that received funds from the Emergency Shelter Grant program in 1992, 14.8 percent came from rental housing and 6.7 percent from owner-occupied homes (Feins & Fosburg, 1998, Exhibit 9). It is not known how many of these people were evicted prior to entering shelter. It is possible that others who came to shelter from the streets or other locations had been evicted previously. But it seems safe to say that targeting people at risk for homelessness solely on the basis of eviction would not be very efficient.
After selecting people at risk for homelessness, based on a more or less sophisticated model, one must then determine what interventions will most readily prevent homelessness, at what cost. The best design for evaluating a prevention program is to randomly assign some proportion of people who meet some risk criteria to receive the specialized program. People who did not receive specialized services would remain free to use other services. Both groups would need to be followed for some reasonably long period of time (years rather than weeks or months) to determine meaningfully what proportion of each group became homeless. If 25 percent of the at-risk group became homeless in the absence of any intervention, and 15 percent became homeless despite the intervention, then one could argue that the intervention prevented 40 percent of the cases of homelessness that would otherwise have occurred [calculated as (.25 - .15)/.25]. Alternatively, one could measure total months homeless in the two groups and determine how many months of homelessness were prevented.
Remarkably few studies of prevention programs have anything approximating this design. Many programs have no comparison group that failed to receive prevention services, much less one that is randomly assigned, and authors make implausible assumptions about the numbers of people who would have become homeless in the absence of intervention (typically assumed to be 100%). Studies frequently have little or no follow-up to determine whether homelessness was prevented, merely postponed, or not affected at all, and often presume success rates of 100 percent for those who received services. Cost-benefit analyses derived from such studies present an illusion of specificity, but, as we will show, different and more plausible assumptions frequently lead to quite different conclusions.
Some observers have likened homelessness to a game of musical chairs in which the players are poor people and the chairs are the housing units they can afford (McChesney, 1990; Sclar, 1990), or in a slightly more sophisticated analogy, the chairs represent the housing poor people can purchase or otherwise occupy by drawing on their personal networks (Koegel, Burnam & Baumohl, 1996). Where there are more poor people than affordable housing units, and where personal networks are attenuated or materially impoverished, some will be left homeless when the music stops. Although individual characteristics may determine who is most vulnerable, and hence be predictive of homelessness, it is resources relative to needs that determine overall prevalence rates (Wright & Rubin, 1991; Koegel et al., 1996). Thus, while homelessness can be prevented by creating resources or reallocating them from those who are not at risk to those who are, reallocation among groups at similar levels of risk is unlikely to affect overall prevalence rates. That is, reallocation affects who gets the housing units, not how many are left homeless when the music stops.
If housing subsidies or other services effectively prevent homelessness for particular individuals, but are in short supply and must be rationed, prevention programs that offer the scarce goods risk reallocating homelessness. Program participants are less likely to become homeless, but those moved back in line or displaced from the queue may be more likely to become homeless. For example, in a sample of families in shelters in New York, two factors predicted receipt of subsidized housing: length of stay in shelter and being assigned to a relatively small, non-profit shelter rather than a congregate shelter or a welfare hotel (Shinn et al., 1988). An earlier analysis of the non-profit shelters suggested that staff did not generate new housing resources; rather, by means of persistent advocacy on behalf of their families, they garnered more than their proportional share of the subsidized housing units available (Shinn, Knickman, Ward, Petrovic & Muth, 1990; for a similar point, see Baumohl & Huebner, 1991). In essence, "months in shelter" reflected families' coming to the top of the housing queue, whereas the success of the non-profit shelters reflected queue-jumping. The overall prevalence of homelessness was not changed by this reallocation of homelessness between those lucky enough to have advocates and those who were not.
Allocation of resources poses a real dilemma for policy makers. Many cities have long waiting lists for public housing. If homeless people are put at the head of the queue, others on the verge of homelessness may be moved back and, as noted, become homeless more often than under a different dispensation. Indeed, if entering shelter is seen as the quickest, most certain route to subsidized housing, shelter entry may be promoted by queue-jumping. (See Culhane, 1992, for a more extended discussion of the perverse incentives created by preferential placements of homeless families.)
This amounts to a cautionary tale for evaluators of programs to prevent homelessness. Even a carefully designed experiment, in which a group randomly assigned to receive preventive services experiences less homelessness than a control group, may not demonstrate prevention (overall reduction in incidence or prevalence) if homelessness has merely been reallocated. At the individual level homelessness has been prevented for program participants, but at the population level, no prevention has occurred. Because overall prevalence rates are very hard to measure accurately, and are influenced by many factors unrelated to the operation of a particular program in a particular area, accurate measures of reductions in the prevalence of homelessness, and unassailable attribution of observed changes to programs, are both unlikely. Rather, we suggest that evaluators consider whether homelessness has been truly prevented or merely reallocated on whatever logical or empirical grounds are available. The reallocation hypothesis is most plausible when the evaluated program involves advocacy for or allocation of existing resources to particular groups. Still, even where the reallocation hypothesis seems persuasive, the program may show that homelessness would truly be prevented if critical resources were more widely available.
The Interagency Council on the Homeless (1994) argued for universal prevention strategies. It noted that for most people, homelessness is a manifestation of extreme poverty, and that ending homelessness will, in the long run, require combating poverty with "more opportunities for decent work, job training that leads somewhere, necessary social services, better education, and affordable housing [all as] components of comprehensive community planning and economic development" (p. 84). It argued against "institutionalizing a separate support system for the homeless population" in favor of improving access to mainstream services (p. 91). Similarly, nearly 4000 providers of homeless assistance, local officials, and homeless and formerly homeless people queried by the Interagency Council rated more affordable housing as the top priority (out of 15 options) for a federal plan to address homelessness (p. 61). (It is not clear how this sample was drawn.)
Jahiel (1992) proposed a variety of universal strategies to prevent homelessness. His recommendations embraced employment (increasing the minimum wage and using the tax system to induce businesses to pay low-income workers more), unemployment and welfare (bolstering incomes, providing vocational rehabilitation and support services such as child care), health (insurance protection for loss of income through illness or disability), education (preparation for job opportunities), and family (attacking underlying causes of family conflict). His most extensive suggestions involved housing. He would increase government support for public housing and non-profit housing corporations (including community-based corporations organized by tenants), preserve existing housing stock by strictly enforcing regulations protecting single-room occupancy hotels (SROs) and other low-income housing, improve municipal services to low-income neighborhoods, and train tenants in methods to resist displacement. He also proposed arrangements to create housing by non-profit corporations or via sweat-equity programs; methods to finance housing via limited-equity cooperatives or master-leasing of shared housing by nonprofit organizations, creation of inexpensive housing such as mobile homes or prefabricated units, and use of non-conventional housing, including SROs (pp. 327-328). Lindblom (1991) offered a similar list, coupled with indicated strategies. Efforts to combat housing discrimination against racial minorities, which remains rampant (Yinger, 1991), might begin to address the over-representation of African Americans in shelter.
Selected prevention strategies might target low-income people who have difficulty affording housing, poor people at particular life stages, or neighborhoods from which large concentrations of homeless people come.
With respect to housing affordability, the Department of Housing and Urban Development (HUD) defines worst-case households as unsubsidized renters with incomes below 50 percent of the area median who pay more than 50 percent of income for housing costs or live in seriously sub-standard housing. These households are at substantial risk of homelessness. One way to estimate the costs of preventing homelessness by attacking housing affordability directly is to calculate the difference between the amount that worst-case households can afford to pay and the actual costs of their units (including rent and utilities other than telephone). The total gap between 50 percent of the incomes of worst-case households and housing costs was $14.3 billion in 1995. If we use the HUD standard that households should pay no more than 30 percent of their income for rent and utilities, the gap between 30 percent of income and housing costs, again for worst-case households, was $22.5 billion in 1995 (figures estimated by Cushing N. Dolbeare from the 1995 American Housing Survey data, personal communication, September 7, 1998).(7) A more generous program to subsidize all households with income less than 50 percent of area median and paying over 30 percent (rather than 50%) of income for rent and utilities would cost more.
These costs are substantial, but they are a fairly small fraction of the tax expenditures that subsidize home ownership, the benefits of which accrue predominantly to wealthier members of society (Dolbeare, 1996). For example, in 1997, homeowners' tax deductions for mortgage interest alone totaled $49.1 billion. If property tax deductions, capital gains deferral, and capital gains exclusions on homes are included, homeowner deductions totaled $90.7 billion (Dolbeare, personal communication).
To put these numbers in further perspective, note that the Interagency Council on the Homeless (1994:85) observed that if the HUD budget simply had increased at the rate of inflation after 1980, budget authority in 1994 would have been $65 billion; HUD's 1994 appropriation was $26 billion. The difference would cover the cost of subsidies to all worst-case households.
There is some evidence that subsidized housing, even without other services, is likely to prevent homelessness for most families. In Philadelphia, the numbers of families admitted to shelter who had been in shelter previously dropped from 50 percent in 1987 to less than 10 percent in 1990 after a policy of placing families in subsidized housing was adopted (Culhane, 1992). Similarly, Wong, Culhane, and Kuhn (1997) found a very low readmission rate (7.6%) among families discharged from shelter in New York City when they received subsidized housing. Shinn et al. (1998) found that New York City families who lived in subsidized housing were less likely to enter shelter in the first place than other families in the public assistance caseload. Further, subsidized housing was very nearly both necessary and sufficient to stabilize formerly homeless families. In a five-year follow-up of a cohort of families who entered shelter, families who received any of several forms of subsidized housing were slightly more likely to have apartments of their own than were a random sample of the public assistance caseload who had never been homeless (97% compared to 92%), and the two groups were equally likely to be stable, defined as having been in one's own apartment without a move for at least a year (80% in both groups). Very few of the formerly homeless families received any services other than subsidized housing (certainly they were not part of special case management programs), yet when they received housing subsidies they attained residential stability. On the other hand, families who did not receive subsidized housing were very unlikely to be stable at the end of five years (38% in own apartment, 18% stable).(8)
Although a variety of factors predicted which families in the public assistance caseload would enter shelter in the first place, only receipt of subsidized housing made any substantial contribution to the prediction of stability at follow-up. Among formerly homeless families, the odds of stability increased twenty fold for households who received housing subsidies, compared to those who did not. Factors that were unrelated to stability, in the context of subsidized housing, included mental illness, substance abuse, health problems, history of incarceration, education, work history, various features of the respondent's childhood (disruptive family experiences, growing up in poverty, teen pregnancy), domestic violence, and strength of personal network, although some of these factors were associated with initial shelter entry (Shinn et al., 1998). Stojanovic, Weitzman, Shinn, Labay, and Williams (1999) found that families (in the same study) who left subsidized housing did so primarily because of serious building problems or safety issues (rats, fire or other disaster, condemnation, or the building's failure to pass the Section 8 inspection).
It is important to note that in New York, families' housing subsidies (and the base rent as well) typically were paid directly to landlords. Thus, families could not delay rent payments to meet other needs. It is not clear whether families would have been as stable five years later if subsidies and base rent payments were more fungible. Lindblom (1996: 193) suggested additional advantages to voluntary programs to provide payments to landlords via an intermediary who could serve as an advocate for tenants' rights: Landlords might negotiate lower rents in exchange for the reliability of cash flow and tenants would obtain more negotiating power because numerous tenants' payments would come through one intermediary.
Less definitive additional evidence that homelessness among families is "cured" by subsidized housing comes from two other studies in which all families received such housing. A nine-city study of homeless families (chosen for long-term patterns of recurrent homelessness and need for services) offered families both subsidized housing (Section 8 certificates) and case management services. Among 601 families on whom 18 months of follow-up data were available, 88 percent remained in permanent housing. This study suggests the value of services-enriched housing and does not speak to the issue of housing without services, although no differences in housing stability were found across sites with rather different configurations of services (Rog, Holupka & McCombs-Thornton, 1995).
Weitzman and Berry (1994) set out to study whether intensive case management services reduced repeat homelessness among New York City families deemed to be at especially high risk on the basis of a set of individual and family risk factors. However, this question could not really be answered because at the end of the one-year follow-up period, the vast majority of families were housed, whether or not they had received the intensive services. Only 8 of 169 high-risk families—just under 5 percent—had returned to shelter. The type of subsidized housing received was the strongest single predictor of who would return, with families in buildings operated by the public housing authority more stable than those in an alternative City program.
Similarly, in a longitudinal study of homeless adults (including a small proportion of women with children) in Alameda County, California, participants were divided into those who attained stable housing, unstable housing, or no housing at a fifteen-month follow-up. Subsidized housing predicted both stably and unstably housed outcomes, and regular entitlement income predicted stable housing; longer histories of homelessness predicted less attainment of either sort of housed outcome; female gender and substance use disorders were associated with unstable housing. Case management services were unrelated to either type of housed outcome (Zlotnick, Robertson & Lahiff, 1999).
Thus, housing subsidy seems to be a very effective preventive measure, but we need more research on different populations in more geographic areas. For individuals with severe mental illness or other disabilities, who are discussed more below, additional services are likely to be necessary, but as there are no studies designed to include assignment to a no-services group, we must point out that this is a commonsensical assertion rather than a demonstrated fact. In any case, there is an important question about whether services should be linked to housing, or whether homeless individuals should make use of services in the community. Culhane (1992:438) notes that providing specialized social services, like providing housing for homeless people only, creates incentives for both policy makers and homeless people to use shelters "as a secondary welfare and housing system."
Other Approaches to Selection. Selected strategies might also target poor people at particular life stages. Studies have consistently shown that homeless families are younger than other poor families (Shinn & Weitzman, 1996). In New York, 53 percent of mothers in families in a cohort entering shelter for the first time were pregnant or had given birth within the past year (Shinn et al., 1998); as noted above, almost half had never had an apartment of their own. [Culhane and colleagues (unpublished papers cited in Culhane & Lee, 1997) found that, over a one-year period, approximately 10 percent of poor children under the age of five in Philadelphia and New York stayed in a public shelter, including 16 percent of poor African-American children.] The cost of starting out in a new apartment (moving costs, first month's rent, security deposit, furnishings) may be prohibitive even for individuals or families who could afford to maintain the housing. A program of loans or assistance directed at first-time renters might permit more young people to make the transition to independent housing, particularly if such a program included work. (We are not aware of any research on such a program.) Assistance to pregnant women and new mothers, beginning with full funding of WIC (the Women, Infant, and Children Food and Nutrition Information Program), might also help young women weather the transition to parenthood.
Another approach would select individuals on the basis of the neighborhoods in which they live. Culhane, Lee, and Wachter (1996) showed that in Philadelphia and New York, between three-fifths and two-thirds of families entering shelter over an extended period came from identifiable clusters of census tracts. The prior addresses of homeless families were more concentrated than the addresses of families in poverty in these cities. In Washington DC, poverty and homelessness were more equally concentrated (Culhane & Lee, 1997). In Philadelphia and New York, rates of shelter admission were strongly related to an area's rates of poor, African-American, and female-headed households with young children and with rates of particularly bad housing conditions. In Washington, rates of female-headed households, female-headed households with preschool children, and unemployed persons were important.
Of course, many of these factors, considered as individual characteristics, also predict entry into shelter, and their design (using census data to characterize neighborhoods with high rates of shelter entry) did not permit the authors to determine to what extent neighborhood characteristics predicted shelter entry above and beyond individual characteristics. The neighborhoods identified were also reasonably large, comprising, in New York, much of the South Bronx, Harlem, and a broad swath of Brooklyn, including Bedford Stuyvesant and East New York. Figures in the article do not permit calculation of the proportions of families in these high-risk areas that entered shelter. Nevertheless, the same types of strategies considered under the rubric of universal prevention could usefully be applied as selected prevention strategies to specific neighborhoods most in need, as judged by the incidence of shelter entry in those neighborhoods. If prevention efforts such as community development, housing construction or rehabilitation, efforts to maintain existing housing stock, job development and training programs, and child care services that permit young mothers to take jobs, are directed at neighborhoods, it is clear that locating services in high-risk neighborhoods makes enormous sense (both from the perspective of preventing homelessness and on other grounds).
Culhane and Lee (1997) noted that families in Washington who requested shelter were put on a waiting list and had to enter shelter to gain access to other services. They suggested that neighborhood-based services brought to families before they enter or even apply for shelter (via indicated strategies triggered by individual needs assessment within the target neighborhoods) might also avert shelter entry for many. We are not aware of any research on the consequences of either selected or indicated neighborhood-based prevention strategies for homelessness, but they are surely worthy of exploration. Those involved in community development initiatives should examine their impact on homelessness.
Despite the Interagency Council's emphasis on universal strategies to alleviate poverty and provide affordable housing, it also advocated indicated prevention methods to 1) "prevent foreclosure or eviction;" 2) "ameliorate domestic conflicts to forestall potentially violent resolutions;" 3) "provide supportive services for physically and/or emotionally disabled individuals;" and 4) "plan for soon-to-be-released inmates in prisons and hospital patients." Surprisingly, without any documentation in an otherwise well-referenced report, the Council suggested that these approaches "are significantly less costly strategies than providing emergency food and shelter for homeless individuals and families" (1994:50-51). Below, we review the evidence for each of these strategies.
Eviction Prevention. Most programs to prevent evictions or foreclosures on mortgages are aimed at families, although single people also get evicted. Typically, these programs offer some combination of cash grants or loans, counsel on budgeting and finances, legal services, mediation or negotiation between residents and landlords or mortgage holders, and advocacy. Often the same agencies also provide secondary prevention services to those already homeless. For example, prevention programs funded by the Emergency Shelter Grants Program (ESG) in Fiscal Year 1991 offered back rent and utility payments (82% of providers), mediation for disputes between landlords and tenants (41%) and legal services for indigent tenants (20%) who faced evictions or utility cutoffs. Many providers also offered payments or loans to families facing foreclosure on their own homes (40%), and security deposits or first month's rent to obtain new housing for people about to be displaced (or, presumably, for people in shelters or shared housing with nowhere to go) (78%). Finally, 25 percent of providers offered referrals and counseling, although it is not clear to what group of clients (Feins et al., 1994a:114).
An evaluation report suggests that "across the entire program, it appears that roughly 205,000 clients and 65,000 families have regained or retained permanent housing through the intervention of the ESG-funded providers" at a cost of about $200 in ESG funds per case (Feins et al., 1994a:186), although the authors also note (p. 206) that it was beyond the scope of the study to assess directly the impact of homelessness prevention activities. The data thus represent agency reports of activities, in one-quarter of cases without any follow-up of the individuals or families helped (Feins, Fosburg & Locke, 1994c:A-91). It is unclear whether any of the agencies corrected their counts for people who later entered shelter or were lost to follow-up, or for those who would have become or remained housed in the absence of intervention. Further, cost estimates may be understated because they include only ESG funds even though the authors suggest that other funds must have been used as well (p. 182).
If these figures are even approximately correct, this is a collection of extraordinarily promising and cost-effective prevention programs, but without more rigorous experimental evaluations, it is hard to credit the results. Descriptions and case studies of individual programs funded under the ESG provide little data on the outcomes of prevention efforts (Feins, Fosburg & Locke, 1994b.)
Schwartz, Devance-Manzini, and Fagan (1991) provided brief descriptions of over 50 state and local efforts to prevent homelessness and longer case studies of seven selected because they were relatively large, established, and well-documented (p. 2). One of the more detailed studies is of a program in Connecticut that provided landlord-tenant mediation and payments of back rent for up to the lesser of two months or $1200 to (then) AFDC families threatened with eviction for non-payment of rent and whose housing was deemed to be habitable, permanent, and affordable. Households were screened and referred to the program by the Department of Human Resources. About half of the cases resulted in mediated agreements between landlords and tenants; surprisingly, in many cases, no financial help from the program was needed. The primary reason for failure, and referral back to the Department of Human Resources, was the client's inability to afford the current rent and secure the tenancy even if back rent were paid (p. 19). The program provided impressive cost-effectiveness figures: In New Haven, the average back rent payment was $960 per family, compared to $7000 for sheltering a family for the allowable maximum of 100 days. In Hartford, 46 families were served at an average payment of $477, compared to $10,514 in shelter costs for 100 days.
Unfortunately, these figures are based on assumptions that we deem implausible. First, the costs of administering the program and mediation were ignored, although in the first months of the program they were substantially higher than the costs of rent payments (p. 15), and the cost of screening families was left out. If we assume that costs of administration and mediation were reduced to equal the costs of back-rent payments after the program was in operation for some time, the estimated costs per family would still need to be doubled. Next, the program assumes that all families who were threatened with eviction would have been evicted, all would have gone to shelter in the absence of the program, and all would have stayed in shelter the maximum of 100 days. As an alternative assumption, if half of those threatened would have been evicted, and half of those evicted would have gone to shelter for 100 days, the cost per shelter episode prevented (including mediation costs) would rise to $3816 in Hartford and $7680 in New Haven, leading to no savings in the latter city. Further, if the average shelter stay were 30 days rather than the maximum of 100 days in Hartford, savings in that city would also evaporate. The authors' calculation also assumes that 100 percent of households who came to a mediated agreement with landlords were prevented from entering shelter. This may be plausible, because 6-month follow-ups were conducted—but the data were not reported.
Thus we see that even simple cost-benefit analyses depend heavily on assumptions that should be put to an empirical test. A more sophisticated analysis might also consider other costs to families who lose their homes and enter shelter (loss of belongings, difficulty in maintaining jobs), costs for stabilizing families after shelter, and benefits to others, such as landlords, when tenancies are secured. All of these factors would enhance the cost effectiveness of the program. In sum, at first glance, the Connecticut program looks promising, but a more rigorous analysis is necessary to determine if it is really cost-effective.
Programs in New Jersey and Maryland provided financial help without extensive mediation or counseling. The New Jersey program gave one-time payments to poor households who could document imminent homelessness or who had already lost housing due to a crisis beyond their control but would be able to afford the housing after the assistance ceased (Schwartz et al., 1991:46). Most payments were to renters, but homeowners facing foreclosure also were eligible for loans. The Maryland program provided rent supplements for up to a year to poor households judged able to return to long-term self-sufficiency in that time. Although "no formal follow-up of clients [was] required or conducted," local administering agencies reported that over one-third of households helped had attained self-sufficiency, and 80-85 percent had attained housing self-sufficiency, although they might receive other entitlements (p. 31). It is not clear what proportion of households were at risk for homelessness or what proportion might have become homeless even with the program.
The New Jersey program conducted one inexpensive follow-up by mailing surveys to 5000 landlords of tenants who had been helped; of those who responded, 72 percent reported that the tenants still resided in the housing or had moved to new situations in good standing with their landlords (p. 84). Thus, 28 percent might be deemed failures. The authors further report that the program cost an average of $1350 per renter household, and that an average stay in shelter in New Jersey lasts 3.5 to 5 months at an average cost of $1500 per month (p. 51) or about four times as much. Again, the cost-benefit calculations required assumptions that could not be tested by evidence. If, in the absence of the program, the failure rate (those leaving in bad standing with landlords) were at least 78 percent rather than 28 percent, and if half of the failures wound up in shelter for an average length of stay, the program would be cost-effective. If the failure rate in the absence of intervention were lower, or the proportion who in fact went to shelter were lower, the program would cease to be cost-effective in the sense of trading program for shelter costs. As in Connecticut, appraising other benefits to tenants and landlords could also alter conclusions.
The HOPE program in Pennsylvania, Ohio, Kentucky, Texas, and Colorado (Schwartz et al., 1991) to prevent mortgage foreclosures provided good evidence that homelessness prevention has benefits to the community that go beyond benefits to people whose homelessness is averted. To households with delinquent mortgages, the program provided intensive financial counseling and support, negotiation with creditors, and various additional resources such as job training, loans, and help obtaining benefits such as energy assistance. Mortgage lenders and utility companies, which stand to benefit when homeowners can pay their bills, provided most of the financing for the program. One utility company estimated that it recouped $9 that would previously have been written off as bad debt for each dollar invested (p. 71). Other programs, discussed more briefly by Schwartz et al. (1991) and Bueno et al. (1989) engaged in novel strategies such as counseling for landlords as well as tenants, coordination among community services, information on entitlements, eviction hotlines, housing clearinghouses and referral networks, and matching services for people who wished to share housing—but offered little evidence of effectiveness.
Even the most sophisticated studies of eviction prevention and rental assistance programs (McIntire, Layzer & Weisberg, 1992; New York State Department of Social Services, 1990) leave much to be desired. In New York, most services were provided to clients on the verge of eviction, and involved mainly legal representation, often combined with advocacy to get clients public benefits. A smaller proportion of programs offered tenant education, counseling, case management, mediation, and tenant organizing. The New York study considered the proportion of households that would avoid eviction in the absence of intervention, and the proportion of cases where evictions were actually averted. It also considered the proportion of evictions that would result in use of shelter, although we believe the figure to be an over-estimate (see footnote 6). Unfortunately, as the report was written after the programs had barely started, the basic evaluation data used were agencies' projections of the total number of clients to be served once the programs became fully operational and estimates of the proportion of closed cases in which eviction would have been prevented or forestalled. In spite of the sophistication of the design, the absence of real program data leads us to discount the estimated cost savings of $4 in shelter costs for every dollar of public investment.
The Washington State study of homelessness prevention and rental assistance programs (McIntire et al., 1992) was one of the few to have short-term follow-up data—at least for four of eleven program sites. The prevention program was aimed at people on the verge of displacement; the rental assistance program served people without housing. Both were cheaper than shelter use (which would have been 2.5 to 2.75 times as much). The authors used the follow-up data to estimate conservatively that 20 percent of those helped might still have gone to shelter, and adjusted cost rates accordingly, but acknowledged that it was difficult to tell how many families would have used shelter without help (p. 86). They suggested that the rate was higher, and hence cost savings may have been greater, for families in the rental assistance program, about half of whom came from shelters or the streets, than for families in the prevention program. Both programs tended to provide case management and sometimes job training services in addition to short-term help with rent or mortgages, and efforts to secure permanent subsidized housing.
In sum, programs to prevent evictions or foreclosures may be of substantial benefit to some households at risk of homelessness and to the communities in which they live. The few studies with follow-up data found that a substantial portion of those who were helped remained housed, at least for the period of assistance, and often appeared to be reasonably stable at the end of that period. But calculation of specific costs and benefits requires data about the extent to which clients of the programs avoid homelessness over the long run and the extent to which they would have become homeless in the absence of the programs. These are rarely collected.
Further, many programs husband their resources by "creaming." That is, they target families deemed most likely to succeed. Many programs serve only households that have sustained sudden losses of income, who can prove they will be able to maintain their residence after receiving help, or who can demonstrate that they are likely to be self-sufficient in the future (see also Lindblom, 1991). Many of the households most likely to become homeless in the absence of the intervention are thus ineligible for current programs. More broadly-based housing subsidies to households with worst-case housing situations would reach a far larger group of those at risk, albeit at greater cost.
Finally, programs to prevent eviction and foreclosure, even if widespread and successful, would reach only a minority of families—those whose homelessness arises from eviction—and would rarely reach single individuals. This limited reach is not a reason to avoid such programs, but suggests that broader action is necessary.
Programs to Ameliorate Domestic Conflicts. Studies consistently find high rates of physical and sexual abuse in childhood, foster care and other out-of-home placements, and/or other family disruptions in the backgrounds of both single individuals and families who enter shelter (Bassuk et al., 1997; Bassuk & Rosenberg, 1988; D'Ercole & Struening, 1990; McChesney, 1987; New York City Commission on the Homeless, 1992; Rog, McCombs-Thornton, Gilbert-Mongelli, Brito & Holupka, 1995; Roman & Wolfe, 1995; Shinn, Knickman & Weitzman, 1991; Sosin, Colson & Grossman, 1988; Susser, Struening & Conover, 1987; Wood et al., 1990). Most studies also find higher rates of domestic violence among homeless than among other poor families (Shinn et al., 1991; Wood et al., 1990), but two studies with more detailed questions (Browne & Bassuk, 1997; Goodman, 1991) found no difference: Rates in both homeless and housed groups were extraordinarily high.
Despite the high rates of family problems and violence in the backgrounds of people who become homeless, it is not clear what a program "to ameliorate domestic conflicts to forestall potentially violent resolutions" (Interagency Council on the Homeless, 1994:50) would look like. Universal strategies to prevent domestic violence and child abuse (by changing norms of acceptable behavior, punishing perpetrators, and providing support and education to parents) and strategies to reduce the need for and increase the quality of foster care would, if successful, reduce these risk factors for homelessness. However, in the cases of child abuse and foster care placement, benefits for the prevention of homelessness would accrue over a very long time. Moreover, as we will note with respect to mental illness and substance abuse, the vast majority of abused and placed children do not become homeless. Designers of indicated interventions for families experiencing domestic conflicts that have not yet become violent face an almost insurmountably difficult task of identifying families to which such interventions would apply. Although programs such as marriage counseling for newlyweds or couples experiencing marital difficulties might well be useful on other grounds, it is quite a stretch to recommend such programs because of their potential to prevent homelessness.
It is even less clear that indicated interventions are advisable to stabilize households already experiencing domestic violence. Service providers report that women are reluctant to leave men who abuse them, in part because of their economic dependence on the men. The need, therefore, is for more, not fewer, shelters, psychological services to traumatized mothers and children, and housing and other resources to help families set up new households. Efforts to get women to stay with perpetrators of violence in order to avoid homelessness would likely lead to injuries and deaths. In a word, they would be misguided. We know of no studies of programs to ameliorate domestic violence as a strategy to prevent homelessness in either the short or the long term, and would hope that anyone who sets one up would look carefully at possible negative consequences.
The fact that childhood abuse and out of home placements, childhood poverty, and adult domestic violence did not detract from the long-term stability of formerly homeless families in New York (once receipt of subsidized housing was accounted for), suggests that these factors may contribute to homelessness largely by restricting housing support of an informal kind. Similarly, the impoverished social ties found in many, but not all, studies of homelessness (see Shinn et al., 1991, for a review) may be important because personal network members can provide or subsidize housing. If housing can be secured by other means (e.g., a government subsidy), it may not be necessary to address underlying problems in relationships or the attenuation of social ties in order to prevent homelessness, though such interventions may be perfectly desirable for other reasons.
Thus, although domestic violence and childhood disruptions may predict homelessness, the best preventive effort may still be access to subsidized housing. Unfortunately, in New York, women who reported domestic violence were somewhat less likely than other women to receive subsidized housing (Shinn et al., 1998). It is unclear whether batterers pursued women into shelters, whether women returned voluntarily to men who abused them, or whether there was some other reason for this pattern.
Supportive Services For Impaired or Disabled Individuals. Popular treatments of homelessness usually emphasize the contributions of one or several major impairments, but the analysts are not over-burdened by their knowledge of epidemiology (Baumohl, 1993). Once the biases of cross-sectional samples, lifetime diagnostic measurements, and other methodological problems are cut away, it is clear that only a minority of homeless single individuals have suffered recently from a major mental disorder, a substance use disorder, or a physical impairment that rises to the level of a work disability—and rates among homeless families are even lower (Koegel et al., 1996; Lehman & Cordray, 1993). Even more important for our purposes, although those with serious impairments are over-represented among homeless people, only a tiny fraction of all people with major physical impairments, mental disorders and/or substance use disorders ever become homeless (Federal Task Force on Homelessness and Severe Mental Illness, 1992; Institute of Medicine, 1990). Thus, although supportive services for people with serious impairments are valuable in their own right, they should generally be justified on grounds other than the prevention of homelessness. From this admittedly narrow perspective, most such services will be wasted.
Among mentally ill individuals, it is not even clear that the most important variables predicting homelessness indicate a lack of supportive services. A project in San Diego examined the relative role of housing subsidies and intensive services for homeless people with severe and chronic mental illness (schizophrenia, bipolar disorder, or major depression). Participants were randomly assigned, in a 2 X 2 design, to access versus no access to Section 8 certificates and to traditional versus comprehensive case management (Hurlburt, Wood & Hough, 1996). Results indicated an enormous effect produced by access to Section 8 certificates. Almost 60 percent of participants with access to the certificates achieved stability in independent housing at the end of the study, compared with 31 percent of participants without access. There was no effect for type of case management on housing outcomes, although all participants received services. Similarly, in the study by Zlotnick et al. (1999) cited above, in which subsidized housing and regular income from entitlements predicted housing stability, but case management did not, about half of the respondents had substance use disorders or dual diagnoses.
A Chicago study of a random sample of both homeless and domiciled individuals with a history of psychiatric hospitalization who obtained their main meal of the day in a free meal program, found that access to material resources was more important than individual characteristics or relationships with the mental health system in predicting homelessness. Income from employment, receipt of Social Security income, and income from welfare all were associated with lower levels of homelessness in a multivariate model. No measure of mental health symptoms or previous hospitalization was significant by itself or in the multivariate model. (Symptoms of alcoholism were measured and not predictive, but symptoms of abuse of other substances were not assessed.) Those currently receiving outpatient treatment were somewhat less likely to be homeless when economic variables were not considered, but outpatient treatment dropped out of the model in the presence of measures of income (Sosin & Grossman, 1991).(9)
These studies suggest that the prevention of homelessness among individuals with serious disabilities, like its prevention among people not so afflicted, should focus on access to subsidized housing and/or to income that allows the individual to rent housing on the open market. Indeed, risk factors for homelessness and protective factors against it among people with serious mental illness [see Lezak & Edgar (1996)] may be significant primarily because they affect a person's access to housing. For example, the difficulty that many people with serious mental illness have in developing and maintaining relationships may reduce the likelihood of obtaining housing and other resources from members of personal networks. If so, two interventions are possible. One could try to attack the problem by bolstering individuals' relationships with families and friends; but a more direct (and arguably more therapeutic) strategy might simply be to provide the housing and other resources that might otherwise come from family and friends. The best protective strategy (housing or income) may not be identical to the risk (attenuated relationships) it counteracts.
Lezak and Edgar (1996) also identified a number of structural risk factors for homelessness which could be altered by changing social policy. These include inadequate discharge planning (see next section), lack of funding and integration for community-based treatment and support services (including community-based crisis alternatives and integrated treatment for mental illness and substance abuse), insufficient disability benefits, and lack of affordable housing or attention to consumer preferences in housing. We focus on the last two, which relate to the need for affordable housing and sustainable sources of income among poor people generally.
Income. Whereas only a small fraction of seriously impaired people become homeless, the low value of Supplemental Security Income (SSI) and General Assistance (GA) benefits virtually guarantees that those who rely on them will have worst-case housing needs. SSI is a means-tested program for disabled, blind, and elderly people with insufficient work histories to qualify for Social Security Disability Insurance, for which basic (non-clinical) eligibility is established through a history of payroll deductions. SSI is thus a welfare program, and in 1990, SSI checks represented only 23 percent of median income, a figure that doubtless is lower in 1999. McCabe, Edgar, Mancuso, King, Ross, and Emery (1993) compared SSI benefit levels to the fair market rent in each county or standard metropolitan statistical area in the United States. On average, renting an efficiency apartment required 66 percent of the SSI check, and renting a one-bedroom required 80 percent. In 9 percent of counties, fair market rent for a one-bedroom apartment exceeded the entire SSI benefit. In the intervening years, the purchasing power of SSI recipients seeking housing in the open market likely has eroded further, as rents almost certainly have risen much faster than benefits.
Federal SSI benefits (sometimes supplemented meagerly by a state) amounted in 1998 to $494 per month for an individual living alone and $741 for a couple living together. These are small amounts of money, to be sure, but they are princely sums by comparison to the benefit levels of General Assistance programs. GA is a generic name for state and local programs that provide ongoing or time-limited assistance to low-income persons who do not qualify for Temporary Assistance for Needy Families (what was AFDC) or SSI—or who are awaiting an eligibility decision by these or other income maintenance programs. Many states do not have GA programs, or GA is operated in only some local jurisdictions; eligibility rules and benefits levels vary dramatically from state to state, or in some states, notably California and Wisconsin, from county to county. Still, as Greenberg and Baumohl (1996:74) point out, "GA programs share one fundamental characteristic: low benefit levels." In 1992, the maximum GA cash benefit for a single adult (the typical recipient), reported by states with uniform statewide programs, ranged from lows of $27 per month in South Carolina and $80 per month in Missouri to highs of $384 per month in Massachusetts and $407 per month in Hawai'i (Burke, 1995:78). Since 1992, GA benefits in many states have declined, eligibility restrictions have been added, some jurisdictions within states have ceased benefits, and the state of Michigan abandoned its GA program altogether (Urban Institute, 1996). In Michigan, 20,000 former GA recipients were evicted following termination of the GA program (Halter, 1996:108).
The miserliness of GA programs, even where they exist, is a major reason why the ranks of the homeless are dominated by unmarried people or couples without children (Burt, 1992: Greenberg & Baumohl, 1996).(10) GA also has particular relevance for impaired people, for many GA recipients suffer with acute and chronic problems that, while making them realistically unemployable, do not meet the stringent Social Security standard of disability (Halter, 1996). Moreover, some impairments, notably substance abuse (since January 1997), do not qualify as the basis for a Social Security disability claim (Baumohl, 1997; Greenberg & Baumohl, 1996).
Housing and Consumer Preferences. Even if modestly augmented by food stamps and Medicaid (which is far from usual), SSI benefit levels simply will not support the configurations of housing and support services desired by impaired consumers and related to residential stability. Tanzman (1993), reviewing studies of mental health consumers' preferences for housing and support services, found that consumers consistently reported that they wanted to live in their own house or apartment. They preferred to live alone or with a spouse or romantic partner, not with other mental health consumers. They wanted staff support that was available, on call, at any time of the day or night, but did not want to live with staff. Finally, they emphasized the importance of material supports, including income, money for deposits on housing, and ongoing financial resources or subsidies for telephones and transportation.
Srebnik, Livingston, Gordon and King (1995) examined the importance of choice in housing among consumers of ten supported housing demonstration projects in five states. Although most consumers had little choice, those who had more, and perceived less influence by others over their choices, were more satisfied and had greater residential stability.
Finally, Carling (1993) reviewed over 4000 journal articles and book chapters on the success of housing programs for people with severe mental illness, including some studies of success in averting homelessness. He concluded that "a comprehensive outreach approach that offers health and mental health services and focuses on the perspectives and demands of clients, work options, and supported housing has been reported to be effective in helping most people overcome homelessness" (p. 440). He questioned the assumption of mental health professionals that people with major mental illnesses need to live in residential programs before living independently, with supports.(11) Comparing individuals with mental illness to poor people, those who are elderly or homeless, and those with developmental disabilities, he suggested that housing needs are similar for each group and that housing problems are less closely related to disability than to economic and social factors such as poverty, shortage of affordable housing, and discrimination. Supports, choices, and control were critical in determining whether people remained in housing, but professionals and consumers in all these groups disagreed about specific needs for housing and supports. Carling judged the supported housing approach, in which assertive community treatment is coupled with independent housing, to be particularly promising for stabilizing mentally ill individuals in the community, although he observed that additional research is needed. Recent evidence for the utility of this approach comes from a study by Tsemberis (1999), who placed mentally ill and dually diagnosed homeless people directly from the streets, drop-in centers, or shelters in their own apartments, with supportive services under consumer control. Participants had higher residential stability than a non-equivalent comparison group in the usual system of graduated residential treatment, involving transitional housing, community residences and supervised SRO hotels, even though moves within the treatment system were not counted against the stability outcome.
In a recent review of clinical research demonstration projects undertaken with Stewart B. McKinney Homeless Assistance Act Funds in five cities, the authors concluded that programs offering a range of housing alternatives, when coupled with case management services, could effectively engage and stably house homeless individuals with severe mental illness (Shern, Felton, Hough, Lehman, Goldfinger, Valencia, Dennis, Straw & Wood, 1997). The experimental manipulation in these studies involved the type and intensity of services offered. Across five cities, between 74 and 88 percent of the experimental groups were in community housing at the final follow-up (which ranged from 12 to 24 months). Excluding data from a sub-study of a street sample in one city, across four cities with data, 78 percent of those in community housing were deemed stable; that is, they had not moved in the last follow-up period. Results were very similar across the diverse interventions. Just as interesting, from 60 percent to 80 percent of the control groups who received less intensive services were also housed in the community. Thus, as noted previously for homeless families, the intensive services made less difference than might have been expected.
Together, these studies and reviews suggest that selected prevention strategies that provided housing subsidies and/or substantially greater welfare benefits to all with worst-case housing needs would be of critical value to people with severe mental illness and those with other impairments as well. In the case of individuals with severe mental illness and/or substance use disorder, aggressive and integrated mental health and substance abuse treatment, combined with housing subsidies and money management services, would seem to be a useful package.
Additional Strategies. Lezak and Edgar (1996) also gave brief descriptions of a variety of specific state prevention programs for people with severe mental illness. These included programs to integrate treatments for people with co-occurring mental illness and substance use disorders, to offer support and training for community living, to enhance discharge planning, to provide crisis services and temporary housing, to increase flexibility in services and funding, to provide comprehensive, integrated systems of care, to create linkages between public housing and mental health agencies, to provide housing options that respond to consumer preferences, to provide culturally competent care, to increase affordable housing options, to provide state rent supplements [described as "one of the most straightforward, effective ways to enable people to afford housing and avoid homelessness" (p. 23)], to supplement SSI grants, to continue rent payments during hospitalization, to develop housing expertise among mental health staff, to reduce the stigma of mental illness, and to support families of people with serious mental illnesses. Additional strategies for the primary and secondary prevention of homelessness among people with severe mental illness and/or substance abuse are currently being evaluated as part of a cooperative agreement funded by the Center for Mental Health Services and the Center for Substance Abuse Treatment (Rickards, Leginski, Randolph, Oakley, Herrel & Gallagher, in press). Strategies include various models for providing housing and services, interventions in which a representative payee helps a consumer to manage money, and family education and respite care. So far, only baseline data have been collected.
Individually and collectively, these strategies represent creative approaches to enhancing quality of life and preventing homelessness among people with severe mental illnesses. Unfortunately, Lezak and Edgar (1996) provide no evidence that any of them actually prevent homelessness. (There "is a need to dedicate more resources to gathering solid data," they observe on page 26.) As is inevitably the case with such syntheses, proposals about what ought to work reflect logical relationships among activities that are extracted from commonsense or some theoretical perspective on good casework. Judged on their own terms, they make sense—but they remain essentially untested.
Discharge Planning. Discharge planning is often recommended as a strategy to prevent homelessness among people being released from institutions. As noted above, a substantial minority of homeless individuals follow institutional circuits, including mental hospitals, jails, and shelters as well as informal housing arrangements and the street, and a smaller minority of both single individuals and parents in families often have foster care placements in their backgrounds. Belcher (1997) documented the costs and problems associated with homeless mentally ill individuals who are repeat users of services and who approach emergency rooms for care. For example, homeless mentally ill people are far more likely than domiciled mentally ill people to enter the criminal justice system and to commit violent crimes (Martell, Rosner & Harmon, 1995; Michaels, Zoloth, Alcabes, Braslow & Safyer, 1992). However, Lindblom (1991) pointed out that relatively few people go directly from institutions to the streets, and there is no evidence that substantial numbers of youths "age out" of foster care with no place to go. Feins and Fosburg (1998, exhibit 9) found that only 14 percent of people in shelters funded by the Emergency Shelter Grants program came from another institution where discharge might have been planned (5.3% prisons or jails, 5.6% detoxification or substance abuse programs, 2% psychiatric facilities, 1.4% residential treatment programs). We have already discussed the limited role of longer-term, supportive services for individuals with severe mental illness in the prevention of homelessness. Discharge planning could be seen as one stage in this process (Lezak & Edgar, 1996) for the subset of mentally ill individuals being discharged from institutions, but there appears to be less research on this aspect of services than on many others.
Belcher (1997) proposes a Community Rehabilitation Program for use at discharge from mental hospitals, with a central role for clinical case managers. The program embodies good casework practice, but the idea that it will prevent homelessness is an article of faith. Indeed, Belcher observes that we know little about the effectiveness of particular interventions. He, too, calls for more research. The Working Conference on Discharge Planning (undated) identified some 30 key components of discharge planning processes, some with as many as four sub-components, organized in the general area of rules and responsibilities, elements of a discharge plan, collaboration and partnerships, and funding. The one we see as incontrovertible recommends research and program evaluation to identify effective services and systems. The role of different plan elements in the prevention of homelessness remains to be demonstrated. Surprisingly, although the plan suggests a starting point for discharge planning (when individuals first enter institutions) it is silent about its duration after people leave. Given the persistent and recurring nature of serious mental illnesses, it seems likely to us that discharge planning may be a good first step, but that more enduring services, especially ongoing housing subsidies, will be necessary to prevent homelessness in the long run.
We are not aware of any experimental evaluations of the efficacy of discharge planning programs per se in preventing homelessness. The study of homeless and domiciled mentally ill individuals in Chicago described above (Sosin & Grossman, 1991) is one of the few to specifically examine the association of discharge planning with homelessness. Among people with histories of psychiatric hospitalization, there was no difference between homeless and domiciled individuals in the percent who had living arrangements made for them at last discharge from the hospital, the percent for whom arrangements involved living with family members, or the percent referred to outpatient treatment at discharge. This was true even before the powerful economic variables were included in the model. Thus, discharge planning, admittedly of a less comprehensive type than proposed by the Working Conference on Discharge Planning, had little impact on homelessness in this sample.
In New York, a study of a "critical time" intervention that was part of the Stewart B. McKinney Demonstration Program mentioned above showed that intensive services offered in the first nine months after men left a psychiatric program in a shelter was associated with a substantial reduction in nights homeless (from 91 to 30) over the 18-month follow-up period (Susser, Valencia, Conover, Felix, Tsai & Wyatt, 1997). The duration of this intervention is substantially longer than what is usually meant by discharge planning. An earlier technical report (Center for Mental Health Services, 1994) noted that housing problems were critical for men in the study, but did not describe the extent to which the intervention helped men to secure stable housing.
In Hines, Illinois, homeless, addicted veterans were assigned to a 12-month case-management program, beginning with approximately three months of residential care, or customary care, beginning with a three-week hospital stay. Differences between the two groups on literal homelessness were substantial at the first three-month follow-up, when many of the experimental group, but none of the comparison group, were still in residential treatment. However, differences diminished thereafter, and reversed direction at the 24-month follow-up (Conrad, Hultman, Pope, Lyons, Baxter, Daghestani, Lisiecki, Elbaum, McCarthy & Manheim, 1998). The authors concluded that "case management needs to be continuous, community-based, and intensive to maintain and/or increase the gains achieved in residential care" (p. 52).
In sum, although discharge planning programs make sense on logical grounds, at least as part of longer-term programs for people with persistent problems, we are unaware of empirical studies of their ability to prevent homelessness, and suspect that more enduring interventions are necessary.
In 1990, the General Accounting Office (GAO, 1990) reviewed what was known about indicated programs to prevent homelessness and concluded that their effectiveness could not be determined because too few collected the necessary follow-up data. Now, nine years later, the same conclusion holds: While a few programs may be promising, none are even near proven. If indicated strategies are to be pursued in the future, we must have more rigorous evaluation designs, including random assignment to treatment and, most important, long-term follow-up of both those in the treatment group and controls. When programs are unable to meet the demand for services, we see no ethical objection to allocating services by lottery among those eligible.
The GAO report did not consider the effectiveness of targeting, but if the goal of prevention is to reduce the incidence or prevalence of homelessness rather than merely to provide useful services to poor people under a politically convenient rubric, targeting is a critical issue. We believe that indicated strategies (e.g., eviction and foreclosure prevention, supportive services for seriously mentally ill people and substance abusers, and discharge planning) will collectively reach only a minority of people who become homeless. Even if they were expanded to reach 100 percent of their intended targets, and were also 100 percent successful in averting the homelessness of those served, they would still prevent fewer than half the annual cases of homelessness. Of course, if an intervention can prevent even a small number of cases of homelessness in an efficient, cost-effective manner, it is a worthy undertaking. But we should at least consider whether broader selected strategies can do better.
Inefficiency is a serious problem with indicated programs. This is because homelessness (even narrowly defined) is not like phenylketonuria: Whereas the latter is an individual, durable, biological trait, the former is the often passing, frequently recurring, complex product of shifting structural influences on individual lives. Homelessness is more the outcome of circumstance—more the product of social contingency—than the predictable fate of certain sorts of poor people. Given this, it should not surprise that individual correlates of homelessness, even when bundled, are inefficient predictors of future homelessness. Indeed, the evidence suggests that it will never be possible to target services sensitively enough to avoid missing a substantial proportion of people who will become homeless or specifically enough to avoid serving several people who will not become homeless for every one who will. To the extent that prevention services are rationed on the basis of individual characteristics, they inevitably will be burdened with the expensive, invidious, and scientifically dubious chore of sorting poor people. Further research on targeting might prove us wrong, but the efficiency of targeting must be demonstrated, not assumed.
Two problems noted by way of introduction are even more fundamental to the practical application of targeting strategies. First, because correlates of homelessness change over time and vary by location, the data on which scientific targeting relies would need to be periodically renewed in the areas to which they are applied. This would be costly, though it would keep a small army of epidemiologists off the street. More troublesome, however, is the inevitable disclosure of targeting criteria to those to whom they apply. A good advocate would do no less, after all, and if a public benefit were at stake, the contents of eligibility criteria would not be protected by law. This would result in relentless manipulation and counter-manipulation between clients and providers, with antagonism and scientific futility as the results.(12)
In view of this assessment, why should we persist with indicated programs heralded as homelessness prevention? There is an old debate about whether material aid and other help for poor people should be narrowly targeted or embedded in universal programs. The argument for targeting emphasizes the tendency of universal programs to "squander" resources on the most privileged; the counter-argument asserts that targeted programs are politically fragile because they alienate middle-income voters (Skocpol, 1991; Wilson, 1987). This larger debate need not concern us here, but there is an analogous question in the prevention of homelessness: Should homeless or "at-risk" poor people get privileged access to resources? The question is important because this is surely what occurs all over the country in the process of queue-forming, whether for subsidized housing in New York City or access to scarce, publicly-funded methadone maintenance slots in San Francisco.
Such preferences reflect moral judgments about relative suffering and culpability, and the relative success of advocates for one group of disadvantaged people or another, but these considerations aside, do they contribute to the efficient prevention of homelessness? In some part, the queue-jumping phenomenon makes such a question difficult to answer because it confounds the efficacy of indicated targeting. Moreover, agency staff sometimes have strong incentives to stretch the official definition of homelessness or risk for it—that is, to widen their nets—and such collusion further complicates the matter. Thus, the street-level politics of categorical distinction and resource rationing (Lipsky, 1980) make it difficult, though not impossible, to rigorously evaluate indicated homelessness prevention activities. We believe that most such programs probably prevent some homelessness, but what they do, for the most part, is help some poor people manage their deprivation a little better.
Broader, arguably more equitable approaches to the prevention of homelessness are essentially speculative, largely untried, and in their own ways, difficult to evaluate. On the surface, at least, they seem expensive, and no more efficient than the indicated programs we have criticized. Even so, the evidence to date suggests, above all, that the most effective levers for homelessness prevention are instruments of housing and income. Writ large in the form of housing, employment, income maintenance, and tax policy, such broad programs would affect the many rather than the few and lift vagrant boats on the flood tide. A selected strategy like subsidies for households with worst-case housing needs [akin to what Skocpol (1991) calls "targeting within universalism"] would not solve the problem of eligibility thresholds that arises in all programs that are not absolutely universal, but it would reach a high proportion of those who become homeless, and would, we believe, markedly decrease homelessness.
Such an approach seems especially urgent in view of HUD data which show that the crisis in affordable housing is worsening even during this period of roaring economic expansion (see footnote 6). Now, in particular, efforts to prevent homelessness must focus on making housing affordable to poor people. Only once this goal is attained does it make sense to consider other objectives. In our opinion, we need at this point to study the impact of saturating several geographically dispersed communities with new Section 8 certificates available to those with worst-case housing needs. (Such projects might be combined with empowerment zones.)
But Section 8 certificate holders need income as well as housing. In the wake of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996, and the collapse of General Assistance, we ought to test selected employment strategies modeled on the Job Corps but modified to include single parents and those with impairments that do not reach the level of a work disability as evaluated by the Social Security Administration. Such programs would address the failures of General Assistance and the homelessness of poor, young parents whose transition from adolescence to adulthood, from family of origin to independent household, historically was aided by welfare and, in recent years, seems to have incorporated frequent (if sometimes brief) shelter utilization.
If universal strategies, or selected strategies directed at abjectly poor people or those with worst-case housing needs, were employed nationwide, evaluation of their discrete contributions to homelessness prevention would be difficult. If they were applied in particular states or communities, evaluation might be possible using time series designs to compare prevalence rates of homelessness in locales with the programs to those without in nearby states or communities subject to the same general economic or social trends. These are the demonstration projects we recommend.
To compare variations on such approaches, housing subsidies and income subsidies, supported work, and public employment could be combined in some places with social services (including representative payee or rent voucher provisions) for substance abusers and people with a serious mental illness. Both services and participants' access to housing and income supports need to be carefully specified, however. As suggested earlier, there is some evidence that direct rent payment may be an important predictor of long-term stability in housing and thus it warrants a separate experimental condition. More generally, it is critical to specify what is meant, precisely, by references to "case management." In the Alameda County study cited above, among homeless adults with major mental illnesses and/or substance use disorder, those with case managers had four to nine times the odds of receiving entitlement income (Zlotnick & Robertson, 1996). In other studies, case management seems to have included social brokerage and advocacy in connection with housing and entitlements, but as its content is essentially unspecified, case management is treated like the proverbial black box. To the extent that it provides access to housing and income, many studies that find significant contributions of case management to housing stability may have obscured the most critical elements of its success.
Finally, Culhane et al.'s (1996) findings on the neighborhoods from which shelter dwellers come suggest the relevance of selected prevention programs that both provide services to individuals and families and utilize community development and community organization methods to enhance the financial, human, and social capital of such immiserated areas.(13) Such programs deserve a test. We do not share Culhane et al.'s faith in the utility of indicated prevention measures within such a selected strategy, but we may be wrong, and certainly there is every reason to believe that community development is vitally necessary if prevention programs are to rise above the mere reallocation of homelessness.
Basler, B. (1985). Koch Limits Using Welfare Hotels. The New York Times, A1, B13, December 17.
Bassuk, E. L. & Rosenberg, L. (1988). Why Does Family Homelessness Occur? A Case-Control Study. American Journal of Public Health, 78(7), 783-788.
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.
Baumohl, J. (1989). Alcohol, Homelessness, and Public Policy. Contemporary Drug Problems, 16: 281-300.
Baumohl, J. (1993). A Dissent from the Manichees. Contemporary Drug Problems, 20: 329-353.
Baumohl, J. (1997). The Elimination of Addiction as a Disabling Impairment in Social Security. Presented at the Center for Substance Abuse Treatment, Rockville, MD, May 29.
Baumohl, J. & Huebner, R. (1991). Alcohol and Other Drug Problems among the Homeless: Research, Practice, and Future Directions. Housing Policy Debate, 2(3): 837-866.
Belcher, J. R. (1997). Discharge Planning. Draft paper prepared for the Working Conference on Discharge Planning, available from National Resource Center on Homelessness and Mental Illness, June 6.
Browne, A. & Bassuk, S. S. (1997). Intimate Violence in the Lives of Homeless and Poor Housed Women: Prevalence and Patterns in an Ethnically Diverse Sample. American Journal of Orthopsychiatry 67: 261-278.
Bueno, I., Parton, M. B., Ramirez, S., & Viederman, D. (1989). When the Rent Comes Due: Breaking the Link Between Eviction and Homelessness: An Eviction Prevention Access Plan. San Francisco: Home Base.
Burke, V. (1995). Cash and Non-Cash Benefits for Persons With Limited Income: Eligibility Rules, Recipient and Expenditure Data, Fiscal Years 1992-1994. Washington, DC: Congressional Research Service.
Burt, M. (1996). Homelessness: Definitions and Counts. In J. Baumohl, ed., Homelessness in America. Phoenix: Oryx Press, 15-23.
Burt, M. R. (1992). Over the Edge: The Growth of Homelessness in the 1980s. New York, Russell Sage Foundation.
Cammaso, M. J., & Jagannathan, R. (1995). Prediction Accuracy of the Washington and Illinois Risk Assessment Instruments: An Application of Receiver Operating Characteristic Curve Analysis. Social Work Research, 19: 174-183.
Carling, P. J. (1993). Housing and Supports for Persons with Mental Illness: Emerging Approaches to Research and Practice. Hospital and Community Psychiatry, 44: 439-449.
Caton, C. L. M., Shrout, P. E., Dominguez, B., Eagle, P. F., Opler, L. A., & Cournos, F. (1995). Risk Factors for Homelessness Among Women with Schizophrenia. American Journal of Public Health, 85: 1153-1156.
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.
Center for Mental Health Services. (1994). Making a Difference: Interim Status Report of the McKinney Demonstration Program for Homeless Adults with Serious Mental Illness. Washington DC: Substance Abuse and Mental Health Services Administration (DHHS # [SMA] 94-3014).
Coleman, J. S. (1988). Social Capital in the Creation of Human Capital. American Journal of Sociology, 94(Supplement), S95-S120.
Conrad, K. J., Hultman, C. I., Pope, A. R., Lyons, J. S., Baxter, W. C., Daghestani, A. N., Lisiecki, J. P., Jr., Elbaum, P. L., McCarthy, M., Jr., & Manheim, L. M. (1998). Case Managed Residential Care for Homeless Addicted Veterans: Results of a True Experiment. Medical Care, 36: 40-53.
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. (1992). The Quandaries of Shelter Reform: An Appraisal of Efforts to "Manage" Homelessness. Social Service Review, 66, 428-440.
Culhane, D. P., & Lee, C. M. (1997). Where Homeless Families Come From: Toward a Prevention-Oriented Approach in Washington, DC. Washington: Fannie Mae Foundation.
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.
Daskal, J. (1998). In Search of Shelter: The Growing Shortage of Affordable Housing. Washington: Center on Budget and Policy Priorities.
D'Ercole, A. & Struening, E. (1990). Victimization Among Homeless Women: Implications for Service Delivery. Journal of Community Psychology, 18: 141-152.
Dolbeare, C. (1996). Housing Policy: A General Consideration. In J. Baumohl, ed., Homelessness in America, Phoenix, AZ: Oryx Press, 34-45.
Edin, K. & Lein, L. (1997). Making Ends Meet: How Single Mothers Survive Welfare and Low-Wage Work. NY: Russell Sage Foundation.
Federal Task Force on Homelessness and Mental Illness. (1992). Outcasts on Main Street: Report of the Federal Task Force on Homelessness and Severe Mental Illness. Washington, DC: Interagency Council on the Homeless.
Feins, J. D. & Fosburg. L. B. (1999). Emergency Shelter and Services: Opening a Front Door to the Continuum of Care. Paper presented at the National Symposium on Homelessness Research, Washington, DC.
Feins, J. D., Fosburg, L. B., & Locke, G. (Abt Associates.) (1994a). Evaluation of the Emergency Shelter Grants Program, Volume I: Findings. Washington DC: U. S. Department of Housing and Urban Development, Office of Policy Development and Research, HUD-PDR -1489.
Feins, J. D., Fosburg, L. B., & Locke, G. (Abt Associates.) (1994b). Evaluation of the Emergency Shelter Grants Program, Volume II: Site Profiles. Washington DC: U. S. Department of Housing and Urban Development, Office of Policy Development and Research, HUD-PDR -1489.
Feins, J. D., Fosburg, L. B., & Locke, G. (Abt Associates.) (1994c). Evaluation of the Emergency Shelter Grants Program, Volume III: Technical Appendices. Washington DC: U. S. Department of Housing and Urban Development, Office of Policy Development and Research, HUD-PDR -1489.
General Accounting Office. (1990). Homelessness: Too Early to Tell What Kinds of Prevention Assistance Work Best. Washington DC: General Accounting Office.
Goodman, L. A. (1991). The Prevalence of Abuse in the Lives of Homeless and Housed Poor Mothers: A Comparison Study. American Journal of Orthopsychiatry, 16: 489-500.
Gordon, R. (1983). An Operational Classification of Disease Prevention. Public Health Reports, 98: 107-109.
Greater Philadelphia Urban Affairs Coalition. (1998). Our Way Home: A Blueprint to End Homelessness in Philadelphia. Philadelphia: Author.
Greenberg, M. & Baumohl, J. (1996). Income Maintenance: Little Help Now, Less on the Way. In J. Baumohl, ed., Homelessness in America: 63-77. Phoenix: Oryx Press.
Halter, A. (1996). State Welfare Reform for Employable General Assistance Recipients: The Facts behind the Assumptions. Social Work, 41(1): 106-110.
Hopper K. & Baumohl, J. (1994). Held in Abeyance: Rethinking Homelessness and Advocacy. American Behavioral Scientist, 37(4): 522-552.
Hopper, K. & Baumohl, J. (1996). Redefining the Cursed Word: A Historical Interpretation of American Homelessness. In J. Baumohl, ed., Homelessness in America. Phoenix: Oryx Press, 3-14.
Hopper, K., Jost, J., Hay, T, Welber, S., & Haugland, G. (1997). Homelessness, Severe Mental Illness, and the Institutional Circuit. Psychiatric Services 48: 659-665.
Hurlburt, M. S., Wood, P. A., & Hough, R. L. (1996). Providing Independent Housing for the Homeless Mentally Ill: A Novel Approach to Evaluating Long-Term Longitudinal Housing Patterns. Journal of Community Psychology, 24(3): 291-310.
Institute of Medicine. (1990). Treating Drug Problems, Volume 1. Washington, DC: National Academy Press.
Interagency Council on the Homeless. (1994). Priority: Home! The Federal Plan to Break the Cycle of Homelessness. Washington, DC: U. S. Department of Housing and Urban Development HUD-1454-CPD.
Jahiel, R. I. (1992). Toward the Prevention of Homelessness. In R. I. Jahiel (Ed.). Homelessness: A prevention-Oriented Approach. Baltimore: Johns Hopkins University Press, 315-334
Knickman, J. R., & Weitzman, B. C. (1989). Forecasting Models to Target Families At High Risk of Homelessness (Final Report: Volume 3). Unpublished report, Health Research Program, Graduate School of Public Administration, New York University.
Knickman, J. R., Weitzman, B. C., Shinn, M., & Marcus, E. H. (1989). A Study of Homeless Families in New York City: Characteristics and Comparisons With Other Public Assistance Families (Final Report: Volume 2). Unpublished report, Health Research Program, Graduate School of Public Administration, New York University.
Koegel P, Burnam M. A., Baumohl J. (1996). The Causes of Homelessness in Homelessness in America. Ed. Jim Baumohl, New York, NY: Oryx Press.
Lang, R. E. & Hornburg, S. P. (1998). What Is Social Capital and Why Is It Important to Social Policy? Housing Policy Debate, 9(1): 1-16.
Lehman, A. F., & Cordray, D. S. (1993). Prevalence of Alcohol, Drug, and Mental Disorders Among the Homeless: One More Time. Contemporary Drug Problems, 20(3): 355-381.
Lezak, A. D. & Edgar, E. (1966). Preventing Homelessness Among People With Serious Mental Illnesses: A Guide for the States. Rockville, MD: Center for Mental Health Services.
Lindblom, E. N. (1991). Toward a Comprehensive Homeless-Prevention Strategy. Housing Policy Debate, 2: 957-1025.
Lindblom, E. N. (1996). Preventing Homelessness. In J. Baumohl (Ed.). Homelessness in America. Phoenix: Oryx Press, 187-200.
Lipsky, M. (1980). Street-Level Bureaucracy. NY: Russell Sage Foundation.
Martell D. A., Rosener, R. & Harmon R. B. (1995). Base-Rate Estimates of Criminal Behavior by Homeless Mentally Ill Persons in New York City. Psychiatric Services 46(6): 596-601
Mashaw, J. L. (1983). Bureaucratic Justice: Managing Social Security Disability Claims. New Haven: Yale University Press.
McCabe, S., Edgar, E. R., Mancuso, L. L., King, D., Ross, E. C., & Emery, B. D. (1993). A National Study of Housing Affordability for Recipients of Supplemental Security Income. Hospital and Community Psychiatry, 44: 494-495.
McChesney, K. Y. (1987). Characteristics of the Residents of Two Inner-City Emergency Shelters for the Homeless. Los Angeles: Social Science Research Institute, University of Southern California.
McChesney, K. Y. (1990). Family Homelessness: A Systemic Problem. Journal of Social Issues, 46(4): 191-205.
McIntire, J. I., Layzer, J., & Weisberg, L. (1992). On Firmer Ground: Housing for Homeless and Near-Homeless Families. Prepared by University of Washington Institute for Public Policy and Management. Olympia, Washington: Washington State Department of Community Development.
Michaels, D., Zoloth, S. R., Alcabes, P., Braslow, C. A., & Safyer, S. (1992). Homelessness and Indicators of Mental Illness Among Inmates in New York City's Correctional System. Hospital and Community Psychiatry, 43: 150-155.
Milofsky, C., Butto, A., Gross, M., & Baumohl, J. (1993). Small Town in Mass Society: Substance Abuse Treatment and Urban-Rural Migration. Contemporary Drug Problems, 20: 433-471.
Mrazek, P. J., & Haggerty, R. J. (Eds.) (1994). Reducing Risks for Mental Disorders: Frontiers for Preventive Intervention Research. Washington: National Academy Press.
New York City Commission on the Homeless. (1992). The Way Home: A New Direction in Social Policy. Report of the New York City Commission on the Homeless.
New York State Department of Social Services. (1990). The Homelessness Prevention Program: Outcomes and Effectiveness. New York, NY: Office of Program Planning, Analysis and Development and Office of Shelter and Supported Housing Programs.
Rickards, L. D., Leginski, W., Randolph, F., Oakley, D., Herrel, J. M., & Gallagher, C. (in press). Cooperative Agreements for CMHS/CSAT Collaborative Program to Prevent Homelessness: An Overview. Alcohol Treatment Quarterly.
Rog, D. J., Holupka, C. S., & McCombs-Thornton, K. L. (1995). Implementation of the Homeless Families Program: 1. Service models and preliminary outcomes. American Journal of Orthopsychiatry, 65, 502-513.
Rog, D. J., McCombs-Thornton, K. L., Gilbert-Mongelli, A. M., Brito, M. C., & Holupka, C. S. (1995). Implementation of the Homeless Families Program: 2. Characteristics, Strengths, and Needs of Participant Families. American Journal of Orthopsychiatry, 65(4), 514-528.
Roman, N. P. & Wolfe, P. B. (1995). Web of Failure: The Relationship between Foster Care and Homelessness. National Alliance to End Homelessness.
Schwartz, D. C., Devance-Manzini, D., & Fagan, T. (1991). Preventing Homelessness: A Study of State and Local Homelessness Prevention Programs. New Brunswick, NJ: American Affordable Housing Institute.
Sclar, E. D. (1990). Homelessness and Housing Policy: A Game of Musical Chairs. American Journal of Public Health, 80: 1039-1040.
Sherman, A., Amey, C., Duffield, B., Ebb, N., & Weinstein, D. (1998). Welfare to What: Early Findings on Family Hardship and Well-Being . Washington, DC: Children’s Defense Fund and National Coalition for the Homeless.
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. & Weitzman, B. C. (1996). Homeless Families are Different. In J. Baumohl (Ed.), Homelessness in America. Phoenix, Arizona: Oryx Press, 109-122
Shinn, M., Knickman, J. R., & Weitzman, B. C. (1991). Social Relations and Vulnerability to Becoming Homeless Among Poor Families. American Psychologist, 46(11): 1180-1187.
Shinn, M., Knickman, J. R., Ward, D., Petrovic, N. L., & Muth, B. J. (1990). Alternative Models for Sheltering Homeless Families. Journal of Social Issues, 46(4), 175-190.
Shinn, M., Weitzman, B. C., Stojanovic, D., Knickman, J. R., Jiminez, L., Duchon, L., & James, S., Krantz, D. H. (1998). Predictors of Homelessness from Shelter Request to Housing Stability among Families in New York City. American Journal of Public Health, 88, 1651-1657.
Skocpol, T. (1991). Targeting Within Universalism: Politically Viable Policies to Combat Poverty in the United States. In C. Jencks & P.E. Peterson, eds., The Urban Underclass. Washington, DC: The Brookings Institution, 411-436.
Snow, D. A. & Anderson, L. (1993). Down on Their Luck: A Study of Homeless People on the Street. Berkeley & Los Angeles, CA: University of California Press.
Sosin, M. R. & Grossman, S. (1991). The Mental Health System and the Etiology of Homelessness: A Comparison Study. Journal of Community Psychology 19: 337-350.
Sosin, M. R., Colson, P., & Grossman, S. (1988). Homelessness in Chicago: Poverty and Pathology, Social Institutions and Social Change. Chicago, IL: University of Chicago, The School of Social Service Administration.
Sosin, M., Piliavin, I., & Westerfelt, H. (1990). Toward a Longitudinal Analysis of Homelessness. Journal of Social Issues, 46(4): 157-174.
Spradley, J. (1970). You Owe Yourself a Drunk: An Ethnography of Urban Nomads. Boston: Little, Brown.
Srebnik, D., Livingston, J., Gordon, L., & King, D. (1995). Housing Choice and Community Success for Individuals With Serious and Persistent Mental Illness. Community Mental Health Journal, 31: 139-152.
Stojanovic, D., Weitzman, B. C., Shinn, M., Labay, L, & Williams, N. P. (1999). Tracing the Path Out of Homelessness: The Housing Patterns of Families After Exiting Shelter. Journal of Community Psychology, 27: 199-208.
Stone, D. ( 1984). The Disabled State. Philadelphia: Temple University Press.
Susser, E. S., Streuning, E. L., & Conover, S. (1987). Childhood Experiences of Homeless Men. American Journal of Psychiatry, 144: 1599-1601.
Susser, E., Valencia, E., Conover, S., Felix, A., Tsai, W., & Wyatt, R.J. (1997). Preventing Recurrent Homelessness Among Mentally Ill Men: A “Critical Time” Intervention After Discharge From a Shelter. American Journal of Public Health 87(2): 256-262.
Swets, J. A. (1973). The Relative Operating Characteristic in Psychology. Science, 182: 990-1000.
Swets, J. A. (1988). Measuring the Accuracy of Diagnostic Systems. Science, 240: 1285-1293.
Tanzman, B. (1993). An Overview of Surveys of Mental Health Consumers' Preferences for Housing and Support Services. Hospital and Community Psychiatry, 44: 450-455.
Tsemberis, S. (1999). From Streets to Homes: An Innovative Approach to Supported Housing for Homeless Adults With Psychiatric Disabilities. Journal of Community Psychology, 27: 225-241.
U. S. Department of Housing and Human Services. (1999). Waiting in vain: An update on America’s Housing Crisis. Available at http://www.huduser/org/publications/affhsg/waiting.
Urban Institute. (1996). Do General Assistance Programs Provide an Adequate Safety Net? Washington, DC: Author.
Weitzman, B. & Berry, C. (1994). Formerly Homeless Families and the Transition to Permanent Housing: High-Risk Families and the Role of Intensive Case Management Services. Final report to the Edna McConnell Clark Foundation. New York University.
Weitzman, B. C., Knickman, J. R., & Shinn, M. (1990). Pathways to Homelessness Among New York City Families. Journal of Social Issues, 46(4): 125-140.
Wilson, W. J. (1987). The Truly Disadvantaged. Chicago: University of Chicago Press.
Wiseman, J. (1970). Stations of the Lost: The Treatment of Skid Row Alcoholics. Englewood Cliffs, NJ: Prentice-Hall.
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.
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.
Working Conference on Discharge Planning. (undated) Exemplary Discharge Planning: Recommendations of The Working Conference. Available from National Resource Center on Homelessness and Mental Illness.
Wright, J. D., & Rubin, B. A. (1991). Is Homelessness a Housing Problem? Housing Policy Debate, 2, 937-956.
Yinger, J. (1991). Housing Discrimination Study: Incidence and Severity of Unfavorable Treatment. Prepared for the U. S. Department of Housing and Urban Development.
Zlotnick, C., & Robertson, M. J. (1996). Sources of Income Among Homeless Adults With Major Mental Disorders or Substance Use Disorders. Psychiatric Services, 47: 147-151.
Zlotnick, C., Robertson, M. J., & Lahiff, M. (1999). Getting off The Streets: Economic Resources and Residential Exits From Homelessness. Journal of Community Psychology, 27(2): 209-224.
NOTES
(1) Put more technically, the solution is not "isomorphic" with the problem.
(2) In 1985, worried that hotel rooms drew people out of sub-standard housing and into the shelter system, New York City made congregate shelters—where scores of families lived in a single, large room with rows of cots—the entry point to the shelter system for families. Said Mayor Koch, "We are going to, whenever we can, put people into congregate housing like the Roberto Clemente shelter—which is not something people might rush into, as opposed to seeking to go into a hotel" (Basler, 1985).
(3) Two of three demographic factors were included in the best model: race/ethnicity (African Americans were at greater risk than Latinos or others), and being pregnant or having an infant under the age of one year. (Youth was related to entering shelter when taken alone, but ceased to be significant when housing factors were entered into the equation, suggesting that youth affected homelessness primarily via access to the housing market.) Two of five measures of human capital or likelihood of remaining persistently poor were included: childhood poverty (defined as receipt of welfare benefits by the family of origin when the respondent was a child) and being married or living with a partner. Surprisingly, marriage increased risk for homelessness. Education, work history, and having been a teen mother were not predictive in the context of other variables. Two measures of disruptive social ties contributed. These were domestic violence in adulthood and family disruption in childhood (a scale that included foster care or other types of separation from the family in childhood or childhood abuse). Positive ties reflected in the respondent's personal network were not predictive in the multivariate model (and at the univariate level, homeless mothers actually had stronger networks than housed mothers; 80 percent had stayed with network members before requesting shelter). Finally, four of five housing factors (doubling up with others, lack of subsidized housing, frequent moves, and overcrowding) predicted shelter entry. Building problems did not because they were almost as severe for poor but housed families as for those who became homeless. None of the four measures of behavioral disorder (mental illness, substance abuse, health problems, or imprisonment) differentiated between homeless and housed