The importance of subsidies in reducing the risk of family homelessness among poor families strongly suggests that increasing the amount and access to these benefits to such families would likely result in a lower incidence rate of family homelessness. Policies to reduce the cost of housing, thereby making it more affordable, are also important. Broad-based efforts to help families pay the cost of housing and to lower such costs are needed to prevent family homelessness.
However, preventing family homelessness in a more targeted fashion by selecting low-income families who are most at risk to be recipients of a preventive intervention remains difficult at this time. This is due both to the difficulty of selecting families who have a very high risk of homelessness and the challenges of ameliorating those risks enough to substantially lower the probability of their becoming homeless.
Trying to broadly identify families who are most vulnerable to homelessness even among extremely low-income people may be inefficient.In the recent reanalysis of the Fragile Families study (Rog et al, 2007), a longitudinal study of a nationally representative birth cohort of new parents and their children, we found that even among women who are extremely poor (at or below 50 percent of the poverty level), the risk of being homeless is not as large as one might expect. Using a very broad definition of homelessness, fewer than 1 in 10 (8 percent) of the women in this poverty sample indicated that they had been homeless for even one night over a one-to three-year period. This number, however, is tempered by the fact that attrition could account for greater difficulty in locating homeless families or families experiencing residential instability at the time the interviews were being conducted. Despite at least 20 telephone or in-person attempts made with each eligible woman (Knab, personal communication), the study lost11 percent of their baseline sample at Year 1, and 14 percent in Year 3, with only 6 percent missing both follow-ups. Approximately half of the missing cases were due to women refusing to be interviewed, being too ill to be interviewed, being incarcerated and unavailable to be interviewed, or being no longer eligible to be interviewed (e.g., parent or focus child was now deceased), while the rest were missing because they could not be located or had moved out of state. If the assumption is made that the group of respondents who could not be located were all homeless, the upper bound for the percentage of women homeless increases to 23 percent for those who are in the poverty sample (Rog and Holupka, unpublished). Attrition analyses also indicated that among the factors that were significant predictors of missing data, several factors, including greater likelihood of reporting baseline substance use and domestic violence and less likelihood of receipt of TANF and/or Food Stamps at baseline, may suggest some level of vulnerability among those missing in Year 3 to being homeless. There are other factors, however, such as one site having greater attrition, that may have more to do with methodology and less with the personal characteristics of the individuals. Even with this upper bound of homelessness, however, fewer than one in four families living at 50 percent of the poverty level or less would be expected to experience homelessness within a three year period.
Similarly, Shinn and colleagues (1998), in their New York City study, needed 20 predictors to distinguish new applicants for shelter from the public assistance caseload in 1988. They were able to build a model that correctly identified 66 percent of shelter entrants while targeting 10 percent of the public assistance caseload. With a large public assistance caseload, however, even 10 percent misidentified as needing services means that four families who would not become homeless would be identified for every family who would receive homeless services; thus 80 percent of services would be wasted. in addition, any model based on a single risk factor would do more poorly and a complicated model such as the one used by Shinn and her associates would be impractical.
Bassuk, Weinreb, Dawson, Perloff, & Buckner (1997), in their multivariate analyses of risk and protective factors that distinguished homeless from low-income housed families, also relied on a number of different variables. Their findings indicate that there are multiple sources of risk for family homelessness (in the realms of mental health, substance use, social supports, housing history and lack of subsidies) and that there is no one standout risk factor that, if ameliorated, would substantially lower the incidence rate of family homelessness.
Thus, targeting families based on their needs, such as domestic violence, mental health, and substance abuse, is likely not to be fruitful given the equally high rates for low-income families generally. In addition, none of these factors predicted shelter entry in New York, when other factors, primarily demographic characteristics and housing histories, were taken into account (Shinn et al., 1998).
Based on the research to date, two groups of families that may be at highest risk are young families and those who have experienced shelter in the past.As noted earlier, studies have consistently shown that homeless families are younger than other low-income families (Shinn & Weitzman, 1996). One possibility of identifying families at risk is to assess the housing assistance needs of pregnant women and mothers of newborns using health clinics serving low-income families. Housing loans and assistance to pregnant and new mothers, such as through WIC (the Women, Infant, and Children Food and Nutrition Information Program) and subsidized child care might help reduce burdens that contribute to financial problems that can lead to homelessness.
In addition, there is a small subgroup of families who return to shelter, even after receiving subsidized housing. In New York, families who left subsidized housing to return to shelter 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) (Stajonovic et al., 1999). Thus, efforts to assure the quality of housing to which families go might lower shelter return rates. Finally, poor families with many competing financial pressures may benefit from subsidies paid directly to landlords, to aid in making housing the first priority.
Another approach to preventing homelessness is to select families on the basis of the neighborhoods in which they live. 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 (Culhane, Lee, & Wachter, 1996). Geographic-based prevention could include a range of environmental- and individual-focused efforts, including housing construction or rehabilitation, job development and training, child care that permits mothers to take jobs, substance abuse treatment, and so forth.