Visible women,invisible gender: Knowledge production on homelessness in Flanders

Introduction

Invisibility is considered to be one of the key aspects of women’s homelessness. Many women are believed to be experiencing the so-called ‘hidden homelessness’: relying on informal support and avoiding homelessness services, thus not being captured in counts (Pleace, 2016). At the same time, an increase in the numbers of women and family homelessness is apparent in many contexts for at least three decades (FEANTSA, 2017). Clearly, there is an inconsistency of alleged invisibility and an increased proportion of women in data collected on populations experiencing homelessness. In this article, I examine the historical roots and current knowledge production on homelessness in Flanders to disentangle some of these contradictions. I argue that while the precarity of women’s housing situations was and is widely acknowledged, gender is not seen as a crucial dimension of vulnerability when it comes to homelessness.

There is a growing body of work on women’s homelessness (Bretherton and Mayock, 2021; Mayock and Bretherton, 2016), however, few studies have explored the processes of gender being made (in)visible through the production of expert knowledge on homelessness in a specific socio-historical context. To date, also, relatively little attention has been paid to the topic of women’s homelessness in Belgium. In an attempt to fill these gaps, this study sets out to explore the following two primary research questions: (1) what are the historical underpinnings of the prevalence and visibility of women’s homelessness in this part of Europe? (2) how are sex and gender represented in the homelessness knowledge production in contemporary Flanders?

Theoretical approach

This article takes on a poststructuralist approach. It recognises social phenomena, such as social problems, as social constructions acknowledging their underlying structures. It follows an assumption that language is central to the knowledge production. There is no universal representation of concepts such as homelessness or gender. Instead, discourses create frameworks that enable and constrain the production of knowledge by permitting certain ways of thinking about reality, while suppressing others. Dominant discourses become rooted in institutions and ways of speaking, writing and representing reality (Hodgson and Standish, 2009). Knowledge and practice are intertwined, for instance, concepts that are used in research, form part of social practice. In policy proposals, selected categories are deployed to produce specific problem representations. Problem representations, in turn, have tangible effects, for example, in the way subjects or target groups of policies are constituted (Bacchi, 1999).

This article explores academic and policymaking knowledge production; especially analyses the attempts to measure the extent of homelessness and categorise groups (notably by gender). This basic information about homelessness (scale and dominant groups) is the cornerstone of any policy design. At the same time, however, such data are produced with a set of assumptions, constrained by methods, and presented in a particular way. In the American context, Jacoy (2012) calls them ‘quantification rituals’ that feed the culture of bureaucracy, while not offering any solutions. Data and numbers, therefore, are also understood here as discourses (Bacchi, 1999, 2017). Furthermore, a wider representation of gender issues in experts’ opinions is analysed. Currently, hegemonic homelessness research in rich countries is presenting a particular view of homelessness as a distinct social problem, often abstracting from the wider structures of inequalities (Farrugia and Gerrard, 2016). New definitions embrace further housing situations and more groups of people experiencing housing precarity; which makes their homelessness ‘more visible’ (Löfstrand and Quilgars, 2016). Yet, it raises the question of whether including women-specific homelessness situations adequately recognises gender as a structural dimension of vulnerability.

Women as a group in homelessness research

In the 1980s and 1990s, some groups ‘re-emerged’ and were framed as ‘the homeless’ (‘shopping bag ladies’ in the States, families, and people from racial and ethnic minorities). Analysing discursive narratives in the American press, Pascale writes of an alienating construction of ‘the homeless’ as a practice of othering people who cannot afford housing to a point that homelessness becomes a ‘master category that overrides gender’ (Pascale, 2005: 258). The ‘discovery’ of new groups challenged the stereotypical ‘middle-aged male alcoholic’ image also in Europe. Since, there is a growing body of research on women’s homelessness (for European reviews, see Bretherton and Mayock, 2021; Mayock and Bretherton, 2016; Reeve, 2018; Savage, 2016). There is also a growing proportion of women in homelessness counts and increased visibility in the public discourse. It is probably due to several intersecting factors – shifting attitudes towards domestic violence, roll-back of welfare states in some contexts, but also the changing understanding of homelessness and the methods of studying it. There is a growing recognition that homelessness means not having a secure, private space over which one can exercise control. As a result, living under a threat of violence is explicitly incorporated into the European Typology of Homelessness and Housing Exclusion (ETHOS; Edgar and Meert, 2005). Women’s specific experiences are thus increasingly included in the homelessness problem. Homelessness is gendered in a complex way, however. It concerns being processed by bureaucracies, criminal justice system, health care system; it relates to relationship breakdown or being alienated from families. Some of these aspects work ‘in favour’ of men and other ‘in favour’ of women, but it is not easy to say in which way it affects numbers that are collected from different sources (see Cramer, 2002: 35–43 for a detailed analysis of ways gender may influence numbers). In addition, policy efforts may concentrate on a problem that is more ‘visible’ (severe) and not necessarily ‘larger’, as with the Rough Sleepers Initiative in 1992 in London which took away financing from women’s initiatives (Abrar et al., 1998).

Moreover, policies, problem representations and research reinforce each other. The concept of invisibility of women’s homelessness is recognised but may be self-reinforcing. For instance, women not only rely on these more ‘hidden’ strategies more often than men, but they may also try to conceal their gender in street settings or avoid certain places for their own safety (Bretherton and Mayock, 2021). In some contexts, many residential services such as crisis centres and domestic violence services are not included in homelessness counts (Baptista, 2010; Mostowska, 2016). Recognition of these mechanisms, however, does not immediately translate into changes in policies, public perceptions, and research. Even if more critical reflections on epistemology and methodology are present (Ryan-Flood and Gill, 2010), the stronger moral judgement is one of the reasons why women’s homelessness is perceived differently than that of men. For example, in research, causes of women’s homelessness are skewed towards passive scenarios, where ‘something happens’ to women, who are then viewed as powerless victims (Bretherton and Mayock, 2021: 24).

In this article, I examine women’s homelessness, its definitions, categorisations, data, and policies as outcomes of a process of knowledge production. In particular, I look at the statistical data as part of the discourses. Enumeration and categorisation are the key elements of the Foucauldian governmentality. Gathering data and classifying it is a crucial component of the modern state, where the numbers and policies are mutually dependent (Dyb, 2021). Social problems are turned into standardised and quantifiable facts. Research, especially producing quantitative data, legitimises both advocacy and government actions. Governing by numbers in homelessness has been extensively explored (Marquardt, 2016; Mostowska, 2020). Jones (2015) demonstrates how the collection of evidence on homelessness in the Reagan Era was negotiated between various stakeholders. For instance, with the change of financing and legislation (the federal Homeless Assistance Act of 1987), point-in-time counts were institutionalised in the United States. It was assumed that numbers were impartial and useful (Jacoy, 2012), leading to the standardisation of data. Only in the late 1990s did longitudinal studies change this perspective and lead to big policy changes (such as housing-led programmes).

Methods

This study is based on the following two principal data sources: (1) an extensive review of historical and current literature, (2) 17 interviews with 22 experts conducted by the author in the years 2019–2021. The research protocol was approved by the ethical commission of the KU Leuven University where the research was carried out (approval number: G-2019 12 1883). The collected material was coded and analysed by the author using NVivo software. Recent literature included scientific articles, policy documents, and reports from counts. They were first reviewed for the most important findings, crucial numbers, and for target groups or priority areas. In these fragments, any mentions of gender or women’s situation or the lack thereof were noted along with the context. Next, the presumed rationale for the selection of these data and its interpretation as it appeared in the document was coded. Apart from the text, the analysis included tables and graphs, and the way data were displayed. The interviewed experts were involved either in policymaking, advocacy, or research in the homelessness or intrafamily violence fields and were based either in Flanders or in Brussels (interviews are numbered continuously I-01 to I-17). Although Flanders and Brussels are two separate regions of Belgium, and Brussels has some unique features in terms of service provision and its own point-in-time homeless count, examples from both regions are included in the analysis. First, because the academic community is to some extent integrated and conducts research in both Brussels and Flanders; second, because a number of institutions are federal in nature or relate to the whole of the Flemish Community and therefore include Brussels. On five occasions, after approaching an institution, an interview was arranged with two representatives. These interviews were not very different from individual interviews, as the experts were mainly speaking on behalf of their organisations. The presence of two interviewees may have contributed to the presentation of more ‘institutional’ views, as opposed to the individual opinions of the interviewees. Experts were asked about their work in general and subsequently directly about their views on the presence of women, and the gender aspect of homelessness in practice, policymaking, and research. Similarly, interview fragments were coded for experts’ explanations regarding the prominence of women’s homelessness (or a lack thereof). These categories were then used for the second iteration of documents’ coding and a subsequent rereading of interviews. The categories that emerged from this analytical process are used in the analytical part of the article.

I will first briefly discuss the Belgian historical background of (1) general state provision and the prevalence and rarity of women’s homelessness; and (2) the historical organisations specifically focused on women. Then I move on to the academic and advocacy knowledge production on homelessness; especially regarding data collected after 2000, to see how sex and gender are represented. I conclude the article with a discussion of the implications of the current state of knowledge.

Historical underpinnings of women’s housing precarity

Housing precarity has been a common phenomenon throughout history. Perhaps that is also why it has not been seen as a distinct social problem. Severe deprivation was prevalent among both men and women, single people and families. Vagrancy was criminalised in the early modern European states, and both men and women were incarcerated and forced to work in workhouses or farm colonies if found not conforming to the norms of settled life (O’Sullivan, 2016). Control and punishment were intensified in the 19th century also in the Low Countries with the advent of the disciplinary society (Maeseele et al., 2014). Women were more likely to be victimised by their family members and those who had ‘no place’ in respectable families included widows, women with children out of wedlock, and girls discharged from orphanages. The Dutch workhouses and farm colonies failed in bringing their inhabitants ‘back to the society’. It was especially difficult for women, who at the time had limited access to housing based on limited ownership rights and inheritance (Van Leeuwen, 2000). Once released from the colonies, they were often voluntarily coming back for lack of other options (Schackmann, 2013: 300). In fact, the revolving door of crime (begging, sex work) and incarceration was the only kind of support that the state had to offer (Van Leeuwen, 2000). At the turn of the 19th and 20th centuries, there were much fewer women than men in the Dutch workhouses for the ‘undeserving poor’. Female vagrancy was associated foremost with sexual conduct but women’s gender-specific vulnerabilities have been largely ignored (Weevers et al., 2012). At the time more of the general poor relief went to the ‘deserving’ women than ‘deserving’ men (Van Leeuwen, 2000).

Charitable institutions of the Catholic church were foremost prominent in Belgium, and they were the first to address women’s situation. ‘Feminisation of charity’ meant that charity care work was increasingly done ‘for women by women’ (Schneider, 2012). It seems thus that women were overrepresented in the population of recipients of charitable support. They were mainly widows and ‘fallen girls’, who were confined to ‘rescue’ or ‘maternity’ homes (Oudshoorn et al., 2018). Women were thus removed from the general, structural poverty, and moved into the women-specific, sexualised problem based on the moral imperative. The division between more ‘disciplinary’ institutions for ‘able-bodied men’ and more ‘caring’ (often religious) institutions for women and children is still visible today.

This widespread destitution ended with the post-war prosperity and the welfare state. Belgian conservative regime adopted many explicitly genderising policies (Saxonberg, 2013). The Belgian Social Pact of 1944 tied social insurance to employment confirming a male-breadwinner model (Pasture, 1993; Van de Broeck, 2002). Institutional childcare for young children was difficult to access, and women were encouraged to take up part-time employment. In addition, the state (and the Catholic church) was promoting suburbanisation and encouraging homeownership (De Vos, 2012). The post-war Belgian welfare state formed thus a dual welfare system – employment-related support was received mostly by men, whereas family assistance by women (Sainsbury, 1999). This nexus of various policies made women especially dependent. In case of a family conflict, their abilities, and resources to form an independent household were severely limited.

Addressing women’s problems specifically came back in the ‘second feminist wave’ which started in Belgium with a women’s strike in 1966. Especially the more radical and revolutionary parts of feminism saw male violence against women as a keystone of male social control and oppression (Van Molle, 2004). Their demands were cutting across different areas of life, questioning the social order as such, and calling for changes in social policy: childcare, family support, abortion, and father’s leaves. In the mid 1970s women’s refuges came into being as grassroots initiatives (De Weerdt, 1980). They were often located in rented apartments, where women shared rooms, cooking and childcare and where they could support each other. Houses were managed in a democratic way with shared responsibilities (Van Eekelen, 1980). Because of a lack of funds and for safety reasons, addresses were changed frequently. Post-war developments have left women more to their own informal strategies or to grassroots ‘women’s’ organisations.

Federalisation and reforms of the Belgian state coincided with the period of deinstitutionalisation (1976) and abolishing of repressive policies such as the ban on vagrancy (1993). In the 1990s, support system was handed over to the local governments and it underwent standardisation. In Flanders, all interventions were eventually integrated into the regional Centres for General Welfare centres (CAWs) in 1997. Charities became secularised and dependent on public subsidies (Dobbelaere, 2010). Women’s refuges were also integrated into this system. Interventions were professionalised, think tanks and other expert advice bodies were established. Some women’s organisations, however, stayed outside social policies and turned into equality bodies.

As in other rich countries, homelessness in Belgium appeared at the time as a distinct social problem. Focus on socio-psychological problems implied that new groups of ‘homeless people’ were ‘discovered’ by the homelessness sector, such as young people outgrowing the youth protection system and ‘battered women’. Since homelessness meant not only a lack of housing but also ‘disaffiliation’, ‘bringing people back to society’ was seen as a goal of these interventions. In studies from the early 2000s, contrasts between men and women were not considered as being structural, produced by policies and services, or being a product of data gathering methods, but rather as individual psycho-social differences. There was a strong focus on individual deficits such as ‘insufficient social participation’ and ‘insufficient normative integration’ (Van Menxel, 2005: 159). It has been found that homeless women were younger and more often taking care of children than homeless men. Women’s problems were identified as being mainly around family relations (Van Regenmortel et al., 2006: 281). It was ‘evident’ that women were reported to shelters with their children, that migrant women were overrepresented, and that there was a clear link between their housing problems, violence, and higher than by men prevalence of psychological problems (Van Menxel, 2005: 168–174). Prevention of violence was named as a principal way to avoid homelessness of women and children. Violence, however, was not seen as caused by structural inequalities between men and women, hence policies focused mainly on support for women already fleeing abusive relationships. There was practically no robust data on the extent of homelessness in Belgium. The research relied on small-scale studies and qualitative methods. Here, women’s life stories did not seem to fit. In a study with 49 participants (of which 10 were female), it was claimed that six interviews did not fit any type – all these life stories were from women (Van Regenmortel et al., 2006: 109–111).

Current knowledge production on homelessness in Flanders

Up until recently ‘numericisation of politics’ hardly occurred in the field of homelessness in Flanders. For decades, there was virtually no data on the extent of homelessness on the national or regional level. Data emerging since the 2010s indicated a large percentage of women and children experiencing homelessness. It was estimated that there were 33% of women in the homeless group in 2002 in comparison with 18% 20 years earlier (Van Menxel, 2005: 166). One of the crucial drivers of change towards producing more reliable numbers was an effort of Flemish researchers taking part in the international community. The EU-funded MPHASIS project (Measuring Progress on Homelessness through Advancing and Strengthening Information Systems) ran in the years 2007–2009 in 20 countries. This transfer of knowledge included defining homelessness and the ways of tallying. Also, looking at various vulnerable groups and available data about them showed the insufficiency of the current administrative data, when homelessness was not the central issue. This has led to several studies commissioned by the Flemish government (Demaerschalk and Hermans, 2010).

A few years later, a ‘base count’ of homelessness in Belgium was attempted (Meys, 2014). It has been a systematic approach to obtaining data which included various categories and a sample of towns and services. It has also pointed to some blind spots, for instance, that the situation of only 15.5% of women in winter shelters demands more comprehensive answers. Overall, in Flanders, there were many women who experienced homelessness: 41% of women among the clients of local welfare offices in the years 2005–2013 and 57.6% among recipients of the social benefit (Meys, 2014). Based on these findings, the Flemish strategy to combat homelessness was published in 2016 (Hermans, 2017).

More studies followed. In quantitative studies, simple comparisons between men and women were presented. Bircan et al. (2017) showed that, even though women were on average younger, they reported more health problems and were less often employed than men. Contrary to a popular image conflating homeless women with single mothers, 59% of women in the study’s sample were single (compared to 86% of men). The majority thus was not accompanied by child(ren). In qualitative studies, even if men and women were compared, gender did not seem to be problematised (De Decker and Segers, 2013; Verstraete et al., 2018).

In short, a significant amount of evidence has been gathered to inform the current knowledge production on homelessness in Flanders. A large proportion of women in previous studies and a body of work concerning specifically women, suggests that the gender dimension is quite salient in homelessness research and policymaking. Here, I look at how gender and sex are represented in the current reports, statistical data and in interviews. First, I start the analysis of the practices of numericisation and the presentation of the sex/gender variable. Sex appears as a variable in most studies; categories being ‘woman’, ‘man’, and sometimes ‘transgender’ or ‘X/non-binary’, thus blurring the boundaries between biological, legal, and gender identity (Lovenduski, 1998). Second, I analyse the discourses around groups of women experiencing homelessness to see how gender is currently represented. Gender is understood here foremost as a sex-based part of social structure that is expressed in relations that are embodied in the sexual division of labour, compulsory heterosexuality, discourses and ideologies of citizenship, motherhood, masculinity, and femininity (Orloff, 1996: 51).

Practices of numericisation

As already mentioned, the numericisation of homelessness knowledge production in Flanders is influenced by international research and by the local context, and recent methods of data gathering and presenting. Thanks to these connections, there is a quick proliferation of evidence-based policies such as Housing First. Producing comparable data is an important part of the current homelessness research. The influential ETHOS typology is selectively used. In the Brussels count, ‘ETHOS is inherited from the past’ along with several Brussels-specific alternations, even though – as one interviewee admitted – ‘some things do not make sense’ (I-16). This pertains mostly to the category ‘inadequate housing’, where many precarious but very different situations are bundled together, such as shelters not recognised by authorities and vacant buildings. Two interviewed researchers noticed that their text about ‘hidden homelessness’ was widely commented on a website for social workers:

[ETHOS] is not a part of a wider knowledge, how broad ETHOS is [. . .] we got a lot of reactions: oh, this is an eye-opener, this is interesting, this is new. (I-11)

The more detailed ETHOS used in Brussels since 2008, had to subsequently be scaled down to the ETHOS-light typology for the data to be comparable with that from other cities (BrussHelp, 2020). The ETHOS-light typology was used because the Scandinavian counts’ methods were considered as the best reference point. The first Belgian city count outside of Brussels was conducted in 2020 in Leuven (Demaerschalk and Hermans, 2020). Other cities and areas followed in 2021 and 2022. The ETHOS-light typology, however, in contrast to the full ETHOS, conflates all shelter-like settings (thus also women’s refuges) into one category and does not include the situation of living under the threat of violence at all. On the ground, gender proved largely irrelevant to most organisations and city officials taking part in the planning of these counts. Even though there was much focus on uncovering ‘hidden homelessness’, it did not indicate women, but rather specific housing situations that the local governments were concerned about (caravan parks, garage boxes, subletting, places with an eviction notice due to safety hazards). The data on these situations were usually not broken down by gender; but more often presented by household type (including a category of ‘single parent’), alternatively, the number of children was reported. At the same time, federal equality bodies such as the Institute for the Equality of Women and Men, admitted that not having such data makes it impossible for them to put the issue on the political agenda (I-13). The same goes for the disconnection between the data on intrafamily violence and housing instability. Still, experts’ opinions were divergent in terms of the usefulness of such count data. Some believed that data are crucial in initiating a political process (I-06), others dismissed such statements claiming that data are easy to manipulate and issues that must be dealt with are much more complex (I-08). The significance of different variables can be anticipated by looking at the ways data are presented. In the 2020 Brussels count, the combination of age and gender was the usual breakdown of groups; hence, percentages for minors, women, men, and transgender people were presented. This way of describing data shifts the focus on the large percentage of children and makes gender much harder to read (BrussHelp, 2020).

According to academic experts, the extent of women’s homelessness in Flanders is not contested. Most often interviewees estimate that there are about 40% of women in the homeless population in Flanders. This figure, as mentioned, comes from the ‘base count’ published in 2014, which is based on older data and is just one of the numbers indicating a percentage of women among the clients of local welfare offices. This shows the common phenomenon of replicating a handy number and the power of the ‘single figure’ (Neylan, 2005). No other convenient number appeared since then. Researchers state that there is little research on the subject. This is caused, in their view, by homelessness research in general not attracting sufficient attention and financing (I-01, I-02, I-07). Also, paradoxically, there is a lack of relevant data that could initiate the process of more data collection (I-14, I-17). As one interviewee put it,

There’s no research but there is a lot of information. No research on the position of women in the system. (I-14)

Strategic areas of prevention and intervention are another example of data being perpetually reproduced in the same categories. Preventing youth homelessness, preventing evictions of families with children, and promoting and implementing Housing First programmes were the priority areas in the Strategy for Preventing and Combating Homelessness from 2016 (Hermans, 2017). The strategy was also based on the data from 2014 and the new strategy, which is being prepared at the time of writing, is likely to repeat these priority areas (I-06). Local programmes are also based on these targets. For instance, in Leuven, the groups that were counted in 2020 are repeated in the new action plan; neither women nor families are named as one of the target groups, even if children are (CAW, 2021). Eventually, such interventions will need to be evaluated; hence, appropriate data on these groups will be collected once again.

Two visible groups of homeless women

As already mentioned, the multilevel federalisation of Belgium contributes to fragmented homelessness policies. In addition, in a multilingual state, many advocacy organisations are divided along language lines and their work is further complicated by the overlapping and unclear competencies of the public administration on different levels (Van den Brandt, 2014). Currently, the municipal Public Centres for Social Welfare (OCMWs) are channelling federal (mostly financial) assistance. The coordination of support on the local level in Flanders is a competence of semi-public Centres for General Welfare (CAWs), which usually apply a case method approach, street work, supported housing and shelters. This statutory support is accompanied by a myriad of non-governmental and informal organisations (Mostowska and Hermans, 2023). In effect, no Belgian or Flemish organisation or agency claims the problem of women’s homelessness or advocates for that group. There are two main areas where women’s homelessness is evident in the discourses: migration and violence.

Among experts working with migrants, the growing share of women in the homeless population is also attributed to an increase in poverty and homelessness among migrants (I-08, I-13). Unlike homelessness, migration is a federal responsibility. This means, that several problems fall outside the ‘homelessness problem’. For instance, the difficult transition from asylum centres to permanent housing for women, especially those who have experienced domestic or sexual violence, was recognised, and addressed, but it was framed as a ‘migration and integration’ issue (I-08). Migrant women are nevertheless present in the mainstream support system. In Antwerp, 70% of women who were reporting to the municipal services, had a migration background. Their main problem, as an interviewee identified, was to find long-term solutions and move on from temporary accommodation:

There’s not a single organization or agency in Antwerp that helps to find housing on the market. [. . .] We see that women with migration background are sometimes isolated for years, never have learnt the language, can’t even take a tram, can’t even draw money from a bank card. If you tell them to go and search for housing . . . that’s perhaps a nice idea but not realistic. (I-09)

Similar programmes (supporting migrants in search of housing on the private rental market) were operating in some municipalities as a ‘buddy’s programme’ for women by women, but were financed as projects and were discontinued because of lack of further funding (I-10). These examples indicate that although migrant women’s homelessness is seen as a distinct problem requiring specific actions, they are not embedded in the mainstream homelessness policy.

Domestic and intimate partner violence is another area associated with women’s homelessness. The Flemish representatives of organisations dealing with ‘intrafamily violence’ did not agree with the ‘gender-based violence’ approach taken mostly by Francophone organisations (I-08). A more ‘Flemish line’ is to see violence as a multidirectional and complex issue affecting the whole family, hence the preferable term ‘intrafamily violence’ (I-08, I-09). From this perspective, women’s homelessness gets conflated with intrafamily violence:

Given that so many women come to shelter because of intrafamily violence, that [violence] is actually embedded in the intake questions. I don’t want to say that we don’t miss people [who experience violence] but it usually means that they want to keep it hidden. (I-09)

Violence appears to the interviewees to be both pervasive and hidden. In the women’s organisations’ perspective, violence is a gendered issue. It was assumed to be a universal reason why women (especially with children) seek refuge. Nevertheless, as already mentioned, the representative of the equality body claimed that there was no ‘hard data’ that the problem of women’s homelessness was caused by violence, so that her agency could act on the grounds of discrimination (I-13). Representative of another ‘women’s organisation’ said they would be interested in any data with a gender perspective:

If we get robust data from others, then we use it to lobby for policy change, right? We’re not a research institution ourselves, but we’ll dig into the latest research, if there’s a strong gender perspective and women are falling out, then we’ll do that. So, we have to produce good data. (I-10)

When speaking about women in general, however, ‘homelessness’ is avoided as a stigmatising term and perhaps a more ‘masculine’ problem, which – when associated with street homelessness – allegedly concerns fewer women. Generally, organisations dealing with ‘women’s issues’ or gender equality talk rather about ‘poverty’ or ‘housing’ (I-10). Other advocacy organisations voiced their concern that different perspectives on women’s homelessness were an obstacle to cooperation between ‘women’s’ and ‘homeless’ organisations (I-03).

The diversity of experiences of homelessness, individualised responses at the local level, the ethnicisation of gender (together with the fact that migration policy is a federal responsibility), seeing violence as a core cause of homelessness and the mutual distancing of organisations all contribute to the fact that women’s homelessness as such is not claimed as a social problem by any particular actor or group, even if both migrant women and women who experiencing violence (and some overlap of these groups that no one speaks explicitly about) are at the centre of women’s homelessness discourse.

For organisations working and advocating for migrants, migration, not gender, is the defining dimension of vulnerability. For women’s organisations, gender meant inequality as manifested in domestic and intimate partner violence. Overall, the concept of ‘gender’ as an element of social structure was rarely used by the interviewees. It was used as a synonym for ‘women’. As one interviewee pointed out, women are not discriminated against by policies (I-06). The individualised responses channelled through the CAWs are both a cause and an effect of the failure to address gender as a structural dimension of vulnerability (‘the state looks at an individual’ I-14). For example, the number of beds available is not disaggregated by gender because most places are offered in single rooms (I-06). Family homelessness and child homelessness were treated as completely separate issues. There was also no reference to the fact that gender inequality also affects men.

Conclusion and discussion

This article has shown the historical context of women’s housing precarity in Flanders, starting in the mid 1800s, through the post-war welfare state, the women’s movement, to the recent development of evidence-based homelessness policies. Although women’s housing precarity was common in the past and is prevalent today, gender is not considered a key dimension of inequality. This situation in Flanders partly has its historical roots. Employment and childcare policies in post-war Belgium were familialising and explicitly genderising. Also, up until recently, ‘numericisation of politics’ hardly occurred in the field of homelessness in Flanders and therefore the extent of women’s homelessness was largely undetermined. The ‘discovery’ of new groups of homeless people, women among them, in the 1990s was coupled with the construction of homelessness as a distinct social problem and subsequently with broadening its definition.

Nevertheless, there is a particular gender blindness in research and practice on homelessness in Flanders. This may be partly a consequence of the complicated Belgian administrative divisions and the complexity of the phenomenon itself. Stakeholders tend to view the problem from a particular policy angle in a local or regional context. Furthermore, the methods and categorisations used to produce quantitative data are constantly caught up in existing knowledge, policies and evaluations, where gender may seem too obvious to be recognised as a crucial characteristic. Where gender could be considered, the problems of defining homelessness and the hierarchy of genders in any given situation are not straightforward, which continues to obscure the gender dimension of homelessness (Cramer, 2002). In the Flemish case, gender is quite invisible when considering practices of numericisation, and ambiguous when analysing discourses on homelessness of migrant women and women experiencing violence.

Not taking the gender aspect into consideration when describing these groups, however, may lead to adverse effects, such as not considering women’s specific needs (in the American context see Zlotnick et al., 2010) or not introducing more trauma-informed policies in settings other than women’s refuges. More research sensitive to the needs of specific subgroups and an intersectional perspective is needed (Zufferey, 2009). Gender invisibility prevails not because women’s homelessness is not acknowledged (it is) but because the gender dimension is not fully applied to what is known about homelessness already (for instance, to what extent one of the most prominent typology of episodic, transient, and chronic homelessness applies to women). Gendered public policy research should address differences between women – and differences between men – as well as between women and men. Feminist researchers warn against using sex as a simple dichotomous variable when we do not consider how masculinities and femininities interact with institutions and policies (Lovenduski, 1998).

Already decades ago, the conceptual shift from sex to gender and the development of theories of gender relations enabled a breakthrough in the study of welfare regimes and the state itself (Orloff, 1996; Sainsbury, 1999). Research on homelessness is still considered young and under-theorised. In this respect, Flanders is probably not unique among rich countries. At the moment, however, knowledge production on homelessness in Flanders mainly reinforces what is already known about women’s homelessness. More gender mainstreaming, participation and representation of different groups could also benefit men, transgender, and non-binary people.