Most people in modern, developed countries live in stable and secure housing. However, at any given time, a small yet consequential fraction of men, women, and children are homeless or live precariously in unstable circumstances. The substantial deprivations of homelessness, especially street homelessness, have made it a focus of social, policy, and research concern. Research across many countries and contexts has improved our knowledge of the incidence, drivers, and consequences of homelessness—the consideration of women’s homelessness has been particularly valuable in this regard. Nevertheless, our understandings remain incomplete, and our policy responses, inadequate. This chapter describes what research tells us about homelessness generally and women’s homelessness specifically. In keeping with the other chapters in this volume, it takes an economic perspective but recognizes the solid scholarly foundation that other academic disciplines have laid. The chapter also describes current policy initiatives to stem homelessness and points to needed research directions.

What Is Homelessness? Is Women’s Homelessness Different?

Although homelessness may seem straightforward to define, there are many definitions with considerable controversy among them and with different implications for women. A common narrow definition considers homelessness as being without housing altogether, a condition that is sometimes referred to as “literal homelessness” or “rooflessness.” The most visible form of this condition involves people (p. 800) who are unsheltered and sleep on sidewalks, in parks, under bridges, in abandoned buildings, or in cars. However, it also includes people who are sheltered in emergency accommodations.

The literal definition has many limitations, especially when considering women’s homelessness. People who lose their houses or apartments might not go directly to the street or an emergency shelter but might instead become part of the “hidden homeless” population by doubling up with friends or family or finding accommodation in unconventional places like rooming houses or hotels. Due to the heightened risks of physical and sexual violence from literal homelessness and sometimes due to their responsibilities for accompanying children, women are much more likely than men to seek out these alternative arrangements. A focus on literal homelessness risks not only overlooking an important set of housing problems but also overlooking women and families.

The limitations of the literal categorization have led researchers to consider broader definitions of homelessness. Mirroring debates from the social construction of poverty, they have advocated for more inclusive definitions of cultural and subjective homelessness. Cultural homelessness refers to living in circumstances that fall below community standards (Chamberlain and MacKenzie 1992). This definition has been operationalized by the Australian Bureau of Statistics (ABS) (Chamberlain and MacKenzie 2008), which distinguishes among primary homelessness (unsheltered literal homelessness), secondary homelessness (staying in various forms of temporary accommodations including friends, relatives, emergency accommodations, and boarding houses), and tertiary homelessness (living in a hotel or boarding house on a long-term basis). Alternatively, subjective homelessness refers to situations that do not meet individual perceptions about the adequacy of a home, including unsafe, insecure, and unstable arrangements (Watson 1984). Subjective definitions have special relevance for women because they incorporate domestic violence and other circumstances that make a home unlivable.

The European Federation of National Associations Working with the Homeless (FEANTSA) has developed the European Typology on Homelessness and Housing Exclusion (ETHOS), which incorporates an even broader set of elements, including some subjective ones (Edgar 2009). The ETHOS identifies four types of homelessness: rooflessness (literal homelessness), houselessness (living in temporary accommodations, longer-term supported accommodations, and institutions), insecure housing (living temporarily with friends or family, under threat or risk of eviction, or at risk of harm), and inadequate housing (living in mobile homes, condemned buildings, or overcrowded circumstances). Through the Homeless Emergency Assistance and Rapid Transition to Housing (HEARTH) Act of 2009, the United States broadened its official definition of homelessness from a narrow literal categorization to include people facing the imminent loss of housing, people fleeing domestic violence, and other situations.

Beyond identifying whether someone is homeless, there are also temporal considerations of how often or how long someone experiences homelessness. In a (p. 801) path-breaking longitudinal study of shelter users in New York and Philadelphia, Kuhn and Culhane (1998) identified three temporal groups: transitionally homeless people who typically experienced a single short stay in a shelter, episodically homeless people who experienced multiple transitions in and out of homelessness, and chronically homeless people who experienced long durations of homelessness and effectively used the shelter system as a long-term source of housing.

Measures and Data Sources

With definitions in hand, it is possible, though challenging, to estimate the number and characteristics of people who are homeless. In a comprehensive methodological review, Edgar (2009) discussed the general sources of homelessness data: person-by-person counts, administrative records, censuses, and broad-based questionnaire surveys.

Counts are the principal method for determining the number of unsheltered homeless people. Each January, community organizations in the United States conduct counts of sheltered and unsheltered people as part of the Annual Homeless Assessment Report (AHAR; see, e.g., HUD 2016). The United Kingdom and several other countries also rely on counts from community organizations, while other countries assign this task to central statistical agencies. There are arguments for each approach. On the one hand, community organizations often know better than central agencies about where local unsheltered homeless people stay. On the other hand, relying on different organizations can lead to inconsistent methods. Local organizations may also lack the funds and expertise to undertake accurate counts. Regardless of who conducts them, the need to make in-person contacts makes counts resource intensive. Because of the time constraints and resource limits, these surveys also tend to collect only a few descriptors about the people being counted and do not follow individual people longitudinally.

Area counts, including the AHAR counts, have been used in research (see, e.g., Early and Olsen 2002; Quigley, Raphael, and Smolensky 2001; Popov 2016). Their chief advantages are their ready availability (they can be downloaded in spreadsheet form) and their variation over time and geography. The main drawback for research analyses is that the data are aggregated and can lead researchers to mistakenly attribute group characteristics to individuals. The sparse set of personal descriptors in the data is also a serious limitation.

Administrative records of people staying in emergency, transitional, or other types of shelters or people applying for or receiving homelessness services are another source of data. The United States uses administrative records from shelters to produce both single-night and annual measures of sheltered homelessness. The United Kingdom and Australia use application and other records from service providers to measure different types of homelessness. Individual-level administrative records, which are generated as a matter of course in assisting the homeless, are inexpensive to examine. They accurately describe how long and how often services are utilized, including the precise start (p. 802) and end dates of usage, and have been an important part of homelessness research (see, e.g., Culhane and Kuhn 1998; Kuhn and Culhane 1998; Wong, Culhane, and Kuhn 1997). Their chief shortcoming is that they are limited to people who use or apply for services and thus may miss people who are unsheltered, doubled up, or in other insecure or inadequate accommodations. Also, the sensitive nature of the data means that they are only available through special arrangements with service providers or government agencies.

A third source of data consists of national censuses and broad-based surveys in the form of self-complete mail-in and online questionnaires and interviewer-administered in-person and telephone questionnaires. These methods, which can be adapted to measure many types of accommodations and personal circumstances, are useful in identifying broader categories of homelessness. Unfortunately, the administration frames for many general or low-income surveys are based on addresses or phone numbers and therefore miss people who are unsheltered, while the administration frames for many homelessness-focused surveys are based on shelter or other service recipients and may miss the nonhomeless population.

An example of the latter type of homelessness-focused survey is the National Survey of Homeless Assistance Providers and Clients (NSHAPC), which was conducted in 1996 with 4,207 clients from a variety of service programs across the United States. Because the sampling frame included soup kitchens, food pantries, health centers, and other services that were not strictly tied to homelessness, the NSHAPC can be used to compare people who were and were not homeless at the time of the interview.

National censuses also attempt to count and collect data on both the housed and homeless populations. Beyond counting people, some censuses, including the US census, only collect information about a few of their characteristics, such as gender, age, nativity, race or ethnicity, and geographic location. Others, including the Australian census, collect a somewhat larger number of descriptors, such as education, disability status, language proficiency, and income, that can be useful in research but are nevertheless still limited in scope.

Rarer are large-scale surveys with housed and homeless people that collect rich sets of measures on people’s characteristics. Two notable exceptions that are greatly adding to our understanding of homelessness are the Fragile Families and Child Well-Being Survey (FFS) in the United States and the Journeys Home: Longitudinal Study of Factors Affecting Housing Stability (JH) survey in Australia. The FFS is a national, longitudinal survey that began with approximately five thousand newborn children who were born in large US cities between April 1998 and August 2000 (McLanahan et al. 2003). The children’s mothers, and where possible fathers, were interviewed when the focal children were born and when the children were ages 1, 3, 5, 9, and 15. Although the FFS was not designed as a homelessness survey, it followed mothers who became homeless, and thus, the survey has become a major resource for research on women’s homelessness (see, e.g., Curtis et al. 2013; Fertig and Reingold 2008; Park, Fertig, and Allison 2011).

In contrast to the FFS, the JH survey was explicitly designed to study homelessness. JH is a national, interviewer-administered survey that began in late 2011 with approximately 1,700 Australian public assistance clients who were either homeless or at high risk of homelessness. It interviewed these subjects in late 2011 and reinterviewed them (p. 803) in five additional waves spaced roughly six months apart, asking about the subjects’ housing situations, family circumstances, physical and mental health, incomes and jobs, substance use, and many other characteristics. The JH survey has been used to study the causes and consequences of homelessness (see Ribar 2017 for a review). Both the FFS and JH survey are available to researchers through license agreements.

As we discuss in subsequent sections, studies of homelessness have also utilized many other smaller or more specialized surveys. These surveys typically rely on convenience samples of shelter or service clients, focus on subsets of the population such as homeless families, or only consider people from a single town or city. However, they have played a valuable role in homelessness research.

The Incidence of Women’s Homelessness

To provide an indication of the incidence and characteristics of women’s homelessness, we review statistics from recent national enumerations. Table 32.1 reports estimates of homelessness from the US single-night AHAR count undertaken in January 2015 (HUD (p. 804) 2016). The United States counted people using the literal definition of homelessness but distinguishing between people who were sheltered and unsheltered. The enumeration also obtained information about each person’s gender, age, race/ethnicity, family status, veteran status, and chronic homelessness.

The estimates show the profound differences in women’s and men’s homelessness. As we mentioned, women make up a minority of the literal homeless population. In 2015, two out of five homeless people in the United States were women, but the ratio increases to four out of nine among sheltered homeless people. Women make up a much smaller minority (28 percent) of homeless individuals living alone; however, women and girls represent the majority (60 percent) of people in homeless families. Females make up half the population of homeless unaccompanied and parenting youths, but they compose only 9 percent of homeless veterans.

To illustrate the differences that alternative definitions and contexts can make, we report homelessness estimates from the 2011 Australian census in Table 32.2. Recall that Australia’s definition extends beyond literal homelessness to include people who double up or live in culturally inappropriate or severely overcrowded accommodations.

As in the United States, there were fewer homeless women than homeless men in Australia in 2011, even with the broader definition. When we look to particular categories, Australia saw half as many unsheltered homeless women as unsheltered homeless men. However, unlike the United States, sheltered homeless women in Australia slightly outnumbered homeless men. The gender ratio among Australians who doubled up was the same as the ratio for homelessness generally. Women were much less likely than men (p. 805) to stay in rooming houses, and roughly equal numbers of Australian women and men lived in temporary lodgings and severely overcrowded accommodations.

Estimates from Europe tell a similar story. Relating figures from the 2011 European census, Busch-Geertsema et al. (2014) reported that women made up a much smaller proportion (15 to 25 percent) of the total homeless population than men across Europe and an even smaller proportion of the unsheltered homeless population. However, women composed a higher proportion of the sheltered homeless population in Sweden and a few other countries. Women also made up a higher proportion of the people qualifying for homelessness services in the United Kingdom.

Besides the point-in-time estimates, surveys have uncovered gender differences in people’s recent and lifetime homelessness experiences. Link et al. (1994) conducted a national phone survey in the United States in which 6.4 percent of women and 8.8 percent of men reported experiencing literal homelessness in their lifetimes. When the definition of homelessness was broadened, 12.8 percent of women and 15.5 percent of men reported either literal homelessness or doubling up during their lifetimes, and 4.1 percent of women and 5.3 percent of men reported experiencing these conditions in the last five years. Chamberlain and Johnson (2015) conducted a similar survey in Australia. Using a broader definition of homelessness, they found higher rates but a narrower difference (11 percent vs. 15 percent) among Australian women and men. However, they also found that women were much less likely than men to have suffered multiple episodes of homelessness, had shorter episodes, and were less likely to have unsheltered episodes. Several studies, including Allgood and Warren (2003), Cobb-Clark et al. (2016), and Culhane and Kuhn (1998), have also analyzed the durations of homelessness episodes and found that women’s episodes are shorter than men’s.

Causes of Women’s Homelessness

Discussions of the causes of homelessness have largely considered three types of explanations: structural factors, personal vulnerabilities, and negative, bad-luck shocks. Structural factors refer to conditions that shape the person’s environment and opportunities and are beyond the person’s control, such as the availability of low-cost or public housing, the level of rents, job market conditions, the amount of social and housing services, and the generosity of welfare payments. Personal vulnerabilities refer to characteristics of the individual that might interfere with finding or maintaining housing. Some of these might be background characteristics, such as a history of abuse, neglect, or deprivation, or ongoing characteristics, such as physical and mental health problems, that are beyond the person’s control. Other characteristics, such as joblessness, heavy drinking, or substance abuse, have some degree of agency. Shocks are unanticipated events, such as a loss of income or welfare benefits, an unexpected medical event or expense, an unexpected end to a lease, or an incident of domestic violence. Research has investigated all three explanations, with studies of structural causes (p. 806) tending to consider homelessness generally but studies of personal vulnerabilities and shocks being more likely to consider women specifically.

Quantitative studies of structural conditions require variation in the structural context, which has led some researchers to examine area- rather than individual-level data. For example, Quigley et al. (2001) and Early and Olsen (2002) considered how numerous structural characteristics might relate to area-level literal homelessness in the United States and found that housing market conditions, such as high rents and the availability of subsidized housing, affected homelessness but that other structural characteristics, such as job market conditions and deinstitutionalization, did not. Unfortunately, their area-level approaches did not account for personal vulnerabilities or shocks and only considered homelessness totals that pooled women and men (but, owing to the focus on literal homelessness, predominantly described men). Of more specific relevance to women, Gould and Williams (2010) examined county-level rates of sheltered family homelessness in Missouri and found that higher unemployment, less generous welfare benefits, and more shelter beds contributed to those rates but that rents did not. The strong association between family homelessness and public provisions of housing and income is consistent with women having greater dependency on these supports given their higher caring responsibilities (Watson 1999). This may be particularly true for homeless young mothers and women who separate in middle or older age (Smith 1999; Waldbrook 2013).

In contrast to these studies, quantitative analyses of personal vulnerabilities have examined individual-level data, which contain descriptors of gender along with other individual background characteristics, allowing researchers to investigate the causes of homelessness specifically for women. Several studies have used case-control samples of homeless and disadvantaged women from within single communities. An influential case-control study by Bassuk et al. (1997) of single-mother families in Worcester, Massachusetts, found that distal personal factors such as childhood foster care placement and a caretaker’s drug use raised the risks of homelessness. It also found that proximal factors such as heavy drinking, drug use, and mental health problems added to the risks, while education, public and social supports, and neighborhood tenure reduced the risks. A weakness of this and other single-community studies is that they lack variation in structural contexts.

More recently, large national and multicity surveys that span multiple geographic contexts, including the NSHAPC, FFS, and JH study, have been used to examine personal and structural vulnerabilities together. Using the FFS, Fertig and Reingold (2008) found that social support, neighborhood tenure, and the availability of affordable apartments decreased family homelessness and that poor health and higher rents increased homelessness. This study was also able to exploit the longitudinal dimension of the data by including a set of controls that were determined before the homelessness spell of interest, therefore addressing concerns with respect to reverse causation. In a similar vein, Johnson et al. (2015), using the JH data, found that housing market conditions predicted entries into homelessness but only for individuals who were not susceptible to recent incarceration, regular drug use, and other risky behaviors. They (p. 807) also found heterogeneous impacts for poor labor market conditions, whereby higher unemployment rates tended to increase homelessness mainly for women and middle-aged individuals.

The third set of explanations for homelessness focuses on negative shocks, particularly among people who are already structurally or personally vulnerable (O’Flaherty 2004, 2009; Lee, Tyler, and Wright 2010). Many of these studies have focused on families. Studies have linked the onset of homelessness with sudden income losses (O’Flaherty 2009), the birth of a child with medical problems (Curtis et al. 2013), and partner incarceration (Geller and Franklin 2014; Wilderman 2014). These latter studies by Curtis et al. (2013), Geller and Franklin (2014), and Wilderman (2014) all exploited the longitudinal nature of the FFS to establish the time ordering of events and minimize reverse-causality issues.

Domestic violence can also be considered to be a shock explanation. Jasinski et al. (2005) interviewed homeless women in the United States and found that one in four reported that violence in their last residence was a cause for their homelessness. Using the Worcester case-control data, Bassuk, Perloff, and Dawson (2001) found that intimate partner violence raised the risks that a baseline group of newly homeless mothers would repeat their homelessness episodes, and Shinn et al. (1998) found that families that entered homeless shelters were more likely to have been exposed to violence than families that remained housed. However, an analysis of first-time entries into homelessness by Lehmann et al. (2007) failed to find significant associations with domestic violence.

The literature assessing the sequences of homelessness or the types of homeless exits, such as episodic and temporary experiences, and homelessness durations have found strong associations between demographic factors, such as age, race, ethnicity, and family structure, and homelessness exit and recurrence. Among different demographic groups, being a homeless family with dependent children was found to have a higher association with the incidence of leaving shelter homelessness, and to have a lower association with the incidence of returning to shelters or the streets (Piliavin et al. 1996; Wong, Culhane, and Kuhn 1997; Culhane and Kuhn 1998). A number of studies were further able to identify discharge locations, and thus distinguish if families left homelessness or simply left shelter accommodation (Stretch and Kreuger 1992; Rocha et al. 1996). Demographic characteristics such as minority racial status (Rocha et al. 1996), the lack of public housing supports (Housing Choice Voucher Program), and sharing lodging with extended families (Stretch and Kreuger 1992) were associated with a lower probability of obtaining permanent housing placement.

The antecedents of other experiences of homelessness, such as doubling up, have been assessed using detailed survey data, such as the FFS and JH (Fertig and Reingold 2008; Curtis et al. 2013; Cobb-Clark et al. 2016). The findings of these studies differ from those assessing the predictors of more conventional types of homelessness. For example, Fertig and Reingold (2008) showed that domestic violence committed against the mother before giving birth and fair market rents were only associated with traditional measures of homelessness. Similarly, Curtis et al. (2013) found that severe infant health (p. 808) conditions were only associated with traditional measures of homelessness. Cobb-Clark et al. (2016) further found that very low levels of education and residing in a nonurban region were associated with lower hazards of exit from literal homelessness but were not associated with a broader measure of homelessness that included doubling up.

Consequences of Women’s Homelessness

The consequences of women’s homelessness are multifaceted and intergenerational. Homeless adult women have poorer outcomes in the domains of health, emotional and general well-being, relationship stability, nutritional outcomes, and economic well-being. As women make up the overwhelming share of heads of homeless families, it is also important to assess the consequences of female homelessness for the welfare of their children.

Theoretical discussions of how homelessness can generate negative outcomes for adults consider the following channels: heightened levels of stress, poorer health, reductions in a person’s market and nonmarket productivity, and acculturation to the homeless subculture. Stress can act as a channel of effect because it can generate further stress and more severe health problems (Pearlin 1989, 1999; Gentzler 2014). Homeless women report enduring and perceiving greater safety issues than homeless men because of their vulnerability to predatory attacks (Milburn and D’Ercole 1991; Jasinski et al. 2005). Poorer health can also result if homeless people have greater exposure to unsanitary physical conditions or highly stressful environments (Lee et al. 2010). Stress and physical and mental health problems can subsequently hamper people’s productivity if they compromise their ability to act rationally (Mullainathan and Shafir 2013) or their capacity to work (Glomm and John 2002). Last, through acculturation or adaptation to the homeless subculture, an episode of homelessness can lessen a person’s odds of escaping homelessness (Culhane and Kuhn 1998; Johnson and Chamberlain 2008). The theoretical discussions of these channels of effect apply to a general homeless population. To the extent that some of these channels are more applicable to women’s homelessness than men’s, we must also consider gendered consequences.

Empirical studies have indeed provided evidence of gendered impacts. Examining the consequences of homelessness for women generally takes one of three forms: a focus on a sample of low-income women (D’Ercole and Struening 1990; Wagner and Perrine 1994; Upshur, Weinreb, and Bharel 2014; Wenzel et al. 2004), a comparison of prevalence rates of various outcomes among samples of homeless men and women (Wenzel, Koegel, and Gelberg 2000; Tyler et al. 2004; Dietz and Wright 2005; Jasinski et al. 2005), and a comparison of housed and homeless persons separately for men and women (Munoz, Crespo, and Perez-Santos 2005; Herault and Ribar 2017; Scutella and Johnson 2017). Among studies that assess the consequences of homelessness for women, many examine women who head a homeless family, which is mainly composed of single mothers.

(p. 809) The literature has primarily examined five consequences of homelessness for women: poorer physical and mental health (including psychological distress), greater physical and sexual abuse, more drug and alcohol problems, increased hunger and food hardships, and lower and less stable employment.

Generally, studies establish a strong link between female homelessness and the consequences of poorer mental health or greater psychological distress (Ingram, Corning, and Schmidt 1996; Munoz et al. 2005; Scutella and Johnson 2017). Often, it is difficult to isolate the impact of homelessness because poorer mental health outcomes can arise from negative life events that predate the woman’s homelessness, such as adverse childhood experiences or recent physical or sexual violence. However, as these studies address some of the concerns of spuriousness, through more comparable control groups (Munoz et al. 2005) or through multivariate regression (Ingram et al. 1996; Scutella and Johnson 2017), we have greater confidence that the heightened levels of psychological distress are likely the outcome of homelessness.

Another consequence of homelessness for women is worse physical health. Munoz et al. (2005) and Wenzel et al. (2004) each compared homeless women with housed women and found that homeless women faced greater risks of physical ailments such as AIDS/HIV. Munoz et al. (2005) further noted the greater prevalence of circulatory problems and skin disorders among homeless women. Interestingly, these physical consequences of homelessness did not appear to afflict homeless men (Munoz et al. 2005). For the outcome of AIDS/HIV, homelessness may have a stronger effect on women because one of the main transmission routes for homeless adults is through heterosexual sexual intercourse (Kerker et al. 2011), and women are more likely to be the victims of rape and sexual assault (Milburn and D’Ercole 1991; Jasinski et al. 2005) but also to barter sex for protection (Johnson, Gronda, and Coutts 2008).

The evidence, however, is more equivocal about other consequences of homelessness. While the association between homelessness and a heightened level of assault and victimization is strong, it remains unclear if assault and victimization are causes or consequences of homelessness. On the one hand, homelessness can facilitate physical and sexual victimization by exposing a person to potential offenders and denying them the privacy and protection of a permanent place of residence (Jasinski et al. 2005; Wenzel et al. 2004). Also, homeless people are often in close physical proximity to each other, increasing the odds of homeless-on-homeless crime (Lee et al. 2010). On the other hand, physical and sexual victimization by a partner is a major risk factor for homelessness, particularly for women (Ingram et al. 1996). Until studies adequately address the issue of reverse causation, it remains unclear if victimization is a consequence of homelessness. Gender differences do, however, appear in the prevalence rates of reported victimization—whereas homeless men were more likely to report physical victimization to the police, women were more likely to report sexual victimization (Jasinski et al. 2005).

Similarly, ambiguities arise in studies that analyze the impact of female homelessness on drug and alcohol abuse. While some studies find that there is a strong positive association (Wagner and Perrine 1994; Munoz et al. 2005; Wenzel et al. 2004), others do (p. 810) not find that homeless women have more drug and alcohol problems than their housed counterparts (Upshur et al. 2014). However, just as reverse causality was an issue for victimization, it also applies to drug and alcohol abuse. This is one area where the general literature on homelessness has progressed further than the female homelessness literature. McVicar, Moschion, and van Ours (2015) detected little effect of homelessness on drug use after they used longitudinal data to consider the temporal ordering of these outcomes and used fixed-effects methods to control for unobservable individual characteristics that might contribute to both homelessness and substance abuse.

Studies of the impact of homelessness on women’s hunger and food insecurity have also produced mixed results. Gundersen et al. (2003) found an association, albeit a weak one, between higher levels of food insecurity and homelessness for lone mothers in data from the Worcester case-control study. However, Herault and Ribar (2017) found that the association was only prevalent for homeless men and not for homeless women in data from the JH survey. Both of these studies are somewhat unique in using instrumental variable methods to address individual episodes of homelessness.

On balance, evidence indicates that homelessness reduces women’s employment. Wagner and Perrine (1994) found that homeless women reported less stable employment than at-risk housed women, and Lei (2013) found that currently homeless people in the NSHAPC were less likely to work in regular jobs than formerly homeless people and that women in these groups worked less than men. Cobb-Clark and Zhu (2017) looked at childhood exposure to homelessness in the JH survey and found that experiencing homelessness at or before age 15 lowered women’s and men’s chances of adult employment. In contrast, however, analyses of the association between more contemporaneous employment measures and homelessness in the JH data by Chigavazira et al. (2014) revealed that homelessness reduced men’s chances of being employed over the next six months but did not affect women’s chances.

Among studies that assess the consequences for women, some focus has been directed at the way in which homelessness impacts their children. As we have discussed, women make up the overwhelming share of heads of homeless families, making the assessment of child outcomes a natural extension of the overall assessment of the consequences of homelessness for women. The process by which a period of homelessness affects a child is likely to have some differences from that which affects an adult. Homelessness can be a uniquely powerful negative life event for a child as it is generally associated with multiple sources of stress, including disruption to school and community relationships, dramatic changes in family routine (Grant et al. 2013), and a general lack of supervision during homeless spells (Haber and Toro 2004). While the impact of homelessness on children is obviously a sizable problem, the risk of obscuring the specific impacts for women is an unwanted side effect.

An extensive literature documents the various consequences for a child’s schooling and academic achievement. Homeless children, compared to low-income housed children, typically experience higher rates of school mobility (Buckner, Bassuk, and Weinreb 2001) and greater academic disengagement (Masten et al. 1993) and, subsequently, have diminished cognitive outcomes (Rescorla, Parker, and Stolley 1991). However, Buckner (p. 811) et al. (2001) and Shinn et al. (2008), who also looked at academic outcomes, found that homeless children have similar academic and cognitive performance as their housed yet low-income peers. The inconsistency in results across these studies with respect to the child’s cognitive performance may arise from the different study groups used and the time span between these investigations where legal changes, such as the Stewart B. McKinney Homeless Assistance Act (of 1987), had been introduced in the interim period.

Turning to health and well-being outcomes, homeless children are shown to be at greater risk of poor nutrition (Molnar, Rath, and Klein 1990), a higher incidence of victimization (McCarthy, Hagan, and Martin 2002), greater exposure to infectious diseases (Haddad et al. 2005), lower access to health care services (Kushel, Vittinghoff, and Haas 2001), and poorer behavioral and health outcomes more generally (Masten et al. 2014). Yet the impact of homelessness on children’s health has been challenged by Park, Fertig, and Allison (2011) who employed a fixed-effects regression to account for observable and unobservable time-invariant confounding factors. Finally, a far smaller literature assesses the long-term consequences for children. As previously mentioned, Cobb-Clark and Zhu (2017) offer a somewhat different focus, finding that childhood experiences of homelessness have implications for adult employment outcomes.

Women’s Coping Strategies

Homelessness is a stressful, detrimental, and dangerous condition, and most people take vigorous actions to avoid it. Should those actions fail, people will take further steps to avoid or minimize its consequences.

As money to pay for housing runs low, people first try to reduce household expenses and may turn to soup kitchens, food pantries, clothing and furniture charities, and other community supports to conserve funds (Sosin 1992). When money runs out or a housing crisis hits, people then approach their personal support networks to find friends and family members with whom they might double up. The existence of these strategies has led some researchers to consider homelessness as a multistage process, with women’s doubling-up with relatives being the “last stage” before descending into homelessness (Liebow 1993, 82; see also Snow and Anderson 1993; Johnson et al. 2008). Consistent with this framing, Wasson and Hill (1998) found that nearly three-quarters of the female-headed families in a sample of New York City shelter entrants came from doubled-up accommodations, while only one-sixth came directly from their own residence. However, the evidence differs across contexts. Data covering all homelessness service requests from Australia in 2013–2014 indicated that 59 percent of women and 42 percent of men initiated homelessness services from their own accommodation, but only 16 percent of women and 15 percent of men initiated services from doubled-up situations (AIHW 2014). Figures from the United Kingdom indicated that 27 percent of the people and families qualifying for homelessness services (mostly women (p. 812) and lone-mother families) came from doubled-up situations (UK Department for Communities and Local Government 2016). The vast differences in these figures suggest that homelessness should be viewed in terms of multiple processes rather than a single process.

Once homeless, people develop a range of survival or coping strategies to fulfill their basic material needs (Snow and Anderson 1987). Several of these activities are common to disadvantaged people generally, such as using meal services (Sosin 1992), drawing on social networks (Meadows-Oliver 2005), scavenging (Snow and Anderson 1987), and engaging in bartering, off-the-books labor, peddling, hustling, and petty crime (Venkatesh 2006). Some of these strategies are gendered. For example, homeless women are much less likely to engage in day labor, peddling, and panhandling than homeless men (Lee and Farrell 2003; Lei 2013). One innovative strategy observed among homeless women in the United Kingdom is the use of public spaces, such as shopping centers, libraries, airports, and public toilets, as places of both safety and convenience (Casey, Goudie, and Reeve 2008). Another gendered behavior is that newly homeless mothers, many of whom are escaping violent partners, devote much energy trying to maintain a semblance of stability in their children’s lives.

Other material coping strategies are concentrated among the longer-term homeless. A particularly risky strategy among women involves moving in with men in exchange for sex, a practice commonly known as “shacking up” (Johnson et al. 2008), as well as prostitution. “Shacking up” enables homeless women to avoid sleeping on the streets where they are much more vulnerable to physical and sexual violence than men. Indeed, many of the coping strategies that women use highlight the acute and specific vulnerabilities they face from homelessness.

In addition to material coping, homeless people undergo different forms of social and psychological coping. One of these involves adapting or reconstructing their identities (Snow and Anderson 1987). Johnson et al. (2008) have described how newly homeless women attempt to “pass as normal” so that they can physically and symbolically distance themselves from the stigma attached to homelessness. Two key strategies are managing information about their situation and developing “disidentifiers” in the form of actions or behaviors, such as staying well dressed, that signal normality. Both strategies are designed to conceal or disguise their homelessness and avoid public and institutional opprobrium about their circumstances.

While many coping approaches help people deal with the coarseness, stress, and insecurity associated with literal homelessness, some can add to psychological stress and perpetuate homelessness (Milburn and D’Ercole 1991). Banyard and Graham-Bermann (1998) reported that homeless women were more likely than other disadvantaged women to adopt unhealthy avoidant behaviors, including bottling up their feelings, refusing to accept their circumstances, and smoking more, but no more likely to adopt healthy active behaviors. As mentioned earlier, the use of alcohol and illicit drugs to escape the brutality and stigma of street life is also a well-documented but detrimental coping strategy among homeless men and women.

(p. 813) The material, social, and psychological strategies that enable people to survive homelessness may form part of a broader process of adaptation and acculturation whereby people become accustomed to and in some ways better at life on the streets (Snow and Anderson 1987). Evidence regarding adaptation, however, is contradictory. Qualitative narratives indicate that homeless people take on adaptive identities (Snow and Anderson 1987) and report greater “street smarts” (Kidd and Davidson 2007) as their homelessness durations increase, but quantitative data from the NSHAPC showed that people’s food hardships grow worse over those durations (Lee and Greif 2008). There are gendered differences in adaptive patterns as well, with men’s levels of psychological distress declining the longer they remain homeless but with women’s levels not declining (Scutella and Johnson 2017).

Men and women share many coping strategies, but there are also differences. For women, what underpins their coping strategies are concerns for children when they are newly homeless, and for those women who remain homeless for longer periods, and who are often single, coping strategies aim to reduce their vulnerability to physical and sexual assault and also to deal with the social stigma attached to being homeless.

Policy Responses

There are many indirect and direct policy approaches to reducing homelessness and alleviating its harms. Indirect policies address general social problems that might contribute to or be exacerbated by homelessness; however, the policies themselves are not framed primarily in terms of homelessness. These include policies to tackle the structural conditions of homelessness, including initiatives to increase the availability of affordable or public housing, strengthen tenant rights, provide more generous welfare payments or basic incomes, and improve job opportunities. They also include policies to attend to personal vulnerabilities, including training to increase job skills, treatment to reduce substance abuse or mental health problems, and protections against domestic and intimate partner violence. Although the policies focus on general social problems, many, including initiatives to stem domestic violence and drug abuse, are also motivated, in part, in terms of their follow-on effects on homelessness.

In contrast, direct policies target homelessness itself or problems specifically among the homeless. The most obvious direct response is to provide emergency accommodation to people who become homeless or who must flee their homes. Governmental and nongovernmental providers operate general homeless shelters that offer protection from the elements and some of the physical dangers of unsheltered homelessness. They also often operate special shelters for women and families escaping domestic violence. Shelters are a common response to homelessness, even if their populations are counted as indicators of the homelessness problem or if they possibly add to homelessness by allowing more families to flee violence or giving homeless people (p. 814) breathing space to search more carefully for permanent accommodations. With respect to these issues, Popov (2016) found that increased funding for a set of homelessness grant programs in the United States, which principally went to shelter provision, reduced the number of unsheltered single adults without adding to the overall number of homeless single adults. The funding similarly lowered the number of unsheltered homeless families but raised the overall number of homeless families. Popov’s results were particularly compelling because he used a historical quirk in the homelessness grant formula as an instrument, or “natural experiment,” for an exogenous shift in grant funding.

To address the other deprivations that homeless people face, governments and community providers offer in-kind assistance, including soup kitchens, food pantries, and drop-in centers. Homelessness agencies also typically offer case management services to help connect people with physical and mental health care, substance abuse treatment, counseling, and other services. In some agencies these services are integrated, but often they are delivered in a fragmented, piecemeal way. Support services can be provided in conjunction with shelters through transitional housing programs that offer short-term accommodations and through supportive housing programs that offer longer-term accommodations. The services in each of these settings attend to immediate needs but also address other problems that act as barriers to people attaining housing.

Another direct policy response is to arrange permanent housing that moves people from being homeless to being stably housed. This is frequently accomplished by prioritizing homeless people’s—and especially homeless families’—eligibility for public housing and housing subsidies and moving them to the front of waiting lists. For people who were chronically homeless and/or had complex needs, the United States and many other countries initially followed a “continuum of care” model that emphasized modifying people’s behaviors and remedying their problems before arranging for permanent housing. The thought was that if these issues weren’t addressed, housing would not be sustainable and people would lapse back into homelessness (see, e.g., Johnson, Parkinson, and Parsell 2012). However, the United States and some other countries have recently shifted to a “Housing First” approach that sees permanent housing as a fundamental need, rather than a contingent one. Housing First policies recognize the value of other supports for chronically homeless people but provide these in conjunction with permanent housing.

Rather than responding after a homelessness episode begins, programs can assist before people lose housing. One prevention strategy is to provide temporary cash payments, housing repairs, legal services, or other aid that helps housed people weather a crisis. Evans, Sullivan, and Wallskog (2016) used day-to-day variation in agencies’ homelessness funding in Chicago to show that prevention payments for rent or security deposits reduced vulnerable families’ entrance into shelters. A notionally simpler prevention strategy is to expand the eligibility for existing homelessness assistance to include people who are at risk of losing housing rather than limiting it to people who have become homeless. A related approach is to extend services to youths and adults who are scheduled to leave institutional care.

(p. 815) Part of the homelessness policy response involves the coordination of services. The causes and circumstances of homelessness are complex, and the needs of people who are homeless or at risk of homelessness often cut across many agency boundaries. As we mentioned, assistance with the structural and personal vulnerabilities is frequently provided through indirect rather than direct programs. These various forms of complexity raise the need for coordination, which can occur by offering them through a single provider. However, effective provision also requires an overarching strategy and coordination that reaches across different programs, government departments, and service organizations. For example, family protection agencies need to coordinate with housing providers. For prevention to be effective, general social service systems and youth and adult institutions need to coordinate with housing agencies.

Another aspect of the policy response is the targeting of scarce resources to those who are viewed as facing the greatest risk of physical and emotional harm. Family homelessness is a priority of many countries’ strategies because of this. Women generally and women experiencing domestic violence are prioritized on similar grounds.

So far, the policy responses that we have discussed are positive in the sense of providing help; however, responses can also be punitive. Some communities have outlawed sleeping outdoors, on sidewalks, or in parks or have closed parks, public buildings, and other places overnight to keep homeless people from sleeping in them. Communities have also criminalized coping behaviors of the homeless, including panhandling, loitering, open drinking, and scavenging.

Many of the positive policy approaches can be seen in the evolution of homelessness policy in the United States. The first major federal legislation in the United States was the 1987 McKinney-Vento Homeless Act, which offered grants for emergency shelters and transitional housing; provided special health care, education, and food assistance for homeless people; created a training program for veterans; and established the US Interagency Council on the Homeless (USICH) to coordinate activities and provide strategic direction. The act was subsequently amended and expanded. For example, new programs were added in 1990 to provide shelter to people with disabilities, mental illness, and health problems and to expand educational services for homeless children, and changes were made in 1992 to address homelessness in rural communities and to consolidate mental health and substance abuse programs.

US homelessness policy was substantially revamped in 2009 through the HEARTH Act. As mentioned, the HEARTH Act broadened the definition of homelessness to include people at imminent risk of homelessness and families and unaccompanied youths living unstably. It expanded homelessness prevention, emphasized rapid rehousing with priority given to homeless families, increased flexibility for rural programs, and changed incentives and selection criteria for funding. In 2010, the United States created the Open Doors strategic plan, which set explicit goals for preventing and ending homelessness among veterans; ending chronic homelessness; preventing and ending homelessness among families, youth, and children; and eventually ending all homelessness (USICH 2015). Open Doors further emphasizes Housing First and rapid rehousing, greater collaboration among and coordination of mainstream services, improved data (p. 816) systems to enhance collaboration, more affordable housing, and coordinated entry systems for housing services that assess housing clients and connect them with the services they need.

What’s Known and What’s Needed to Know More

As this chapter has shown, the general social problem of homelessness affects both women and men, yet there are many aspects of homelessness that are unique to women. Overall, women are much less likely than men to find themselves homeless, and if they become homeless, women tend to have shorter spells. However, differences in the proportions of women and men who are homeless narrow when we consider particular categorizations, such as sheltered homelessness and doubled-up situations. Also, the circumstances of women’s homelessness differ from men’s, with women being vastly overrepresented in family as opposed to single-adult homelessness.

Several causes of homelessness are also highly gendered. This is especially true of domestic and intimate partner violence, which many women report as a precipitating reason for their homelessness. Women are susceptible to other adverse family circumstances, including family breakdown, children’s medical problems, and even partner incarceration. At the same time, disadvantaged women are affected by several of the same vulnerabilities as men, including high rents, shortages of affordable housing, and childhood backgrounds with abuse, neglect, or deprivation.

There are also gender differences in the consequences of homelessness. Homeless women tend to report and experience greater safety issues and sexual victimization than homeless men. The associations between homelessness and physical ailments such as AIDS/HIV are also stronger for women than for men. Another consequence of homelessness that is highly gendered is the intergenerational impact. As women make up a much higher share of heads of homeless families than men, the impacts of homelessness on children are distinctively a mark of women’s homelessness. Children of homeless women are found to have poorer health and well-being outcomes, and some were found to exhibit lower achievement in school and for cognitive tests.

Women’s coping strategies when confronted with a housing crisis or with homelessness differ from men’s, with some strategies advantaging women but with others disadvantaging them. In terms of the helpful strategies, women appear to take special steps to draw on their social networks and double up when faced with the loss of their own accommodations; this helps them avoid literal homelessness. However, other strategies are more detrimental. Once homeless, women seem more reluctant than men to pursue the underground economic strategies of day labor or panhandling, but some women also adopt the highly risky strategies of sexual barter and prostitution.

(p. 817) Similarly, women are advantaged by some policy responses but not by others. Female-headed families are prioritized for emergency homelessness services and for housing placements. Special shelters and housing programs for victims of domestic violence also principally help women. In contrast, services that target individual unsheltered homeless adults, the chronic homeless, and veterans benefit women less because of their underrepresentation in these groups.

To further analyze these issues, we need to improve the data that are collected. Extant research relies heavily on cross-sectional data, which provides only a static picture of homelessness and is therefore of limited use in the creation of social policy. Indeed, in recognition of the limitations of cross-sectional data, researchers have long argued that longitudinal data are critical for the development of meaningful policy and program responses.

Longitudinal data on homelessness have been around since the early 1990s when the first longitudinal studies were conducted in the United States (Sosin, Piliavin, and Westerfelt 1990). Longitudinal studies have helped us to better understand homelessness as a process of exclusion and adaptation that takes place over time and that affects men and women in different ways. However, most longitudinal studies rely on relatively small samples of already homeless people, drawn from a single geographic area. Further, they typically track people for a relatively short period of time. To advance our understanding of the dynamics of homelessness, large-scale longitudinal studies that include both at-risk and homeless households, especially individual women, are required. Future longitudinal studies should also seek to draw samples from multiple locations. This enables researchers to exploit variation in housing and labor market conditions and to investigate the impact of different public policies in areas. Further, there is considerable value in collecting detailed housing information between waves. Such data enable researchers to more accurately investigate the timing of homeless spells and the factors that precede them. Satisfying these conditions is not particularly easy and can be expensive, but the Australian JH survey provides a model.

While large-scale longitudinal studies can provide data that can help address key policy questions, there are opportunities to go further. Around the world, governments are collecting more and more administrative data. Integrating administrative data from housing, welfare, justice, health, and other allied social services, as well as developing data linkages with longitudinal panel datasets, could yield important information about patterns of service use and service outcomes. While administrative data have limitations (they only capture people who use services, the breadth of information measured can be narrow, and data quality for some nonservice characteristics can be poor), integrating administrative and longitudinal datasets offers new opportunities to design targeted interventions to different groups of at-risk and homeless persons. Similarly, studies of homelessness have not fully exploited mixed-method approaches. Well-conducted mixed-method studies that add qualitative narratives to other quantitative surveys and administrative data can yield more nuanced insights into the factors that prevent homelessness and the factors that drive entries into and exits from homelessness.

(p. 818) In addition to the need for better data, we see that much of the evidence regarding the causes of, consequences from, and interventions for homelessness is associational. There is a critical need for more rigorous analyses. Within the past few years, a few innovative quasi-experimental studies have been undertaken, most notably the evaluation of community homelessness grants by Popov (2016) and prevention payments by Evans et al. (2016). The quirk in the homelessness grant formula that Popov uncovered and the day-to-day variation in agencies’ assistance funding that Evans et al. detected could be applied to study other outcomes, including women’s homelessness. Rigorous evaluations would also include random-control-trial analyses of programs and formal cost–benefit analyses.

The lack of longitudinal data and the infrequent use of experimental designs means that we still do not know a great deal about the dynamics of women’s homelessness and the impacts of interventions designed to both prevent and end homelessness. Indeed, it is a key reason that we know so little about the ways that family dynamics influence women’s pathways into and out of homelessness. The paucity of information on family dynamics spills over into the area of children’s homelessness. Most homeless children are in the care of their mothers, and until we can better understand why some women with children end up homeless while others do not, developing appropriate policy responses remains a challenge. While better data can help to illuminate the factors that put families at risk of homelessness, as well as the factors that impede their exits from homelessness, future investigations need to consider the structural and systemic contexts in which such risks emerge. Indeed, in recent times the risk of homelessness among older women appears to have increased. Why older women are suddenly at greater risk than in the past can only be understood by examining the characteristics of older women and how they interact with contemporary social and economic conditions.

Author’s Note

The findings and views reported in this chapter are those of the authors and should not be attributed to the Melbourne Institute of Applied Economic and Social Research or RMIT University.