OECD Toolkit to Combat Homelessness

2. Measurement: Definitions, data and drivers

Abstract

Block 2 discusses the critical role of robust data collection in creating the evidence base for homelessness policies. Reliable data, grounded in a clear, consistent definition of homelessness, enables policy makers to monitor trends, allocate resources effectively, and develop evidence‑based strategies. In practice, measuring homelessness is fraught with methodological challenges, making cross-country comparison difficult. This block outlines ways to strengthen data collection approaches and improve the quality and coverage of homelessness data.

Relevance and key data

Effective homelessness policies rely on a regular assessment of the scale, drivers, and scope of the challenge. A consistent statistical definition and standardised data collection methods for homelessness enable policy makers to monitor trends, strategically target interventions, optimise resource allocation and make evidence‑based decisions (Busch-Geertsema, Culhane and Fitzpatrick, 2016[1]; Hermans, 2023[2]). Moreover, data collection can enhance public service delivery, foster collaboration through shared understanding, spark innovation and bolster government accountability (van Ooijen, Ubaldi and Welby, 2019[3]). However, data collection should always be considered as a means towards sounder homelessness policies, rather than an end in and of itself (Hermans, 2023[2]).

In practice, measuring homelessness is fraught with methodological challenges (Drilling et al., 2020[4]), making cross-country comparison of homelessness statistics difficult (OECD, 2024[5]).

Methodological challenges to measuring homelessness

A number of methodological challenges and limitations stymy a comprehensive assessment of the state of homelessness in OECD and EU countries, and render cross-country comparison difficult. This section summarises the discussion of methodological issues and proposed solutions to strengthen data collection on homelessness that are developed in further detail in the OECD Monitoring Framework: Measuring Homelessness in OECD and EU countries (OECD, Forthcoming[6]) and the OECD Affordable Housing Database (indicator HC3.1) (OECD, 2024[7]).

A complex, dynamic phenomenon that is hard to measure

As research has shown, homelessness is a complex phenomenon, resulting from both “macro” and “micro” circumstances, and affecting an increasingly heterogeneous population (Lee, Shinn and Culhane, 2021[8]). Some individuals experience a short period of homelessness or housing instability, while others – generally a smaller share of the population – experience prolonged periods of homelessness or transition in and out of homelessness over the course of several weeks, months or years (i.e. “chronic homelessness”), and may have higher social support needs (OECD, 2015[9]; OECD, 2020[10]). The dynamic nature of homelessness can be hard to capture in data collection exercises: street counts, for instance, which are generally conducted at a specific point-in-time, thus “miss” individuals who are not experiencing homelessness at the moment of the count. This diversity is also reflected in the different pathways into homelessness that have been observed across countries (Box 2.1).

Box 2.1. There are many different pathways into homelessness

Research has pointed to structural factors, institutional and systemic failures and individual circumstances (or a combination of these) that may contribute to an individual becoming homeless (see (OECD, 2020[10]). Depending on the country, some factors may be more or less relevant:

• Structural factors include tight housing market conditions, labour market changes, poverty, a shrinking social safety net, migration policies, or reductions in housing allowances. Research has identified a correlation between homelessness and rising housing costs; other studies have pointed to a link between homelessness levels and increasing aggregate poverty rates (Baptista and Marlier, 2019[11]; Quigley, Raphael and Smolensky, 2002[12]). In the United States, the shortage of affordable and social housing has been found to be a major factor in contributing to homelessness, particularly in high-cost urban areas (Colburn and Page Aldern, 2022[13]).

• Institutional and systemic failures refer to the higher risk of housing instability among people transitioning out of institutional settings (such as youth or foster care, the criminal justice system, the military, or hospitals and mental health facilities). In France, for instance, around one in four homeless adults born in the country was previously in foster care or known to child welfare services (FAP, 2019[14]). In Canada, research has found that youth experiencing homelessness are much more likely to have been involved with the child welfare system than the general public (Gaetz et al., 2016[15]); a national youth homelessness survey conducted in 2019 revealed that more than seven in ten people who first experienced homelessness before the age of 16 had a record with child protection services (Bonakdar et al., 2023[16]). Other drivers can include racism (see, for example, (Olivet et al., 2021[17]; Paul et al., 2019[18])) and sexism (for a discussion of the gendered dimensions of homelessness see (Bretherton, 2017[19])), homophobia, and transphobia (see, for instance, (Ecker, Aubry and Sylvestre, 2019[20])).

• Individual circumstances, including traumatic events, such an eviction or job loss, a personal crisis (family break-up or intimate partner violence), child poverty, and health issues (mental health or addiction challenges) are also correlated with homelessness (see, for instance, (Johnson et al., 2015[21]; Ministry of Housing, 2019[22]; Piat et al., 2015[23]). Intimate partner violence (IPV) is a leading cause of homelessness among women (OECD, 2023[24]; Sullivan et al., 2023[25]).

Absence of a harmonised definition of homelessness, resulting in considerable cross-country differences in what it means to be homeless

There is no internationally agreed definition of homelessness, and countries’ statistical definitions vary considerably. In the European Union, many countries use the ETHOS Light framework, which aims to provide a “common language” for assessing and comparing different types of living situations of people experiencing homelessness (Table 2.1). Most official homelessness statistics at national level cover ETHOS Light categories 1, 2, and 3 (rough sleepers and people staying in emergency or temporary accommodation). A smaller share of official statistics include ETHOS Light 4, 5, and 6 (people staying in institutions, living in non-conventional dwellings, or doubling up with family and friends) (cf. the OECD Affordable Housing Database (OECD, 2024[7]) and the OECD Monitoring Framework (OECD, Forthcoming[6])). Further, some national homelessness statistics cover living situations that go beyond the ETHOS Light typology. In many OECD countries outside the EU, homelessness data can be aligned to a large extent with the ETHOS Light categorisation; however, in some cases – such as the Netherlands, the Slovak Republic and the United Kingdom – it is not as straightforward.

Table 2.1. Harmonising the definition of homelessness: ETHOS Light typology

	Operational category	Living situation	Definition
1	People living rough	Public spaces/external spaces	Living in the streets or public spaces without a shelter that can be defined as living quarters
2	People in emergency accommodation	Overnight shelters	People with no place of usual residence who move frequently between various types of accommodation
3	People living in accommodation for the homeless	Homeless hostels Temporary accommodation Transitional supported accommodation Women’s shelters or refuge accommodation	Where the period of stay is time‑limited, and no long-term housing is provided
4	People living in institutions	Health care institutions Penal institutions	Stay longer than needed due to lack of housing; no housing available prior to release
5	People living in non-conventional dwellings due to lack of housing	Mobile homes Non-conventional buildings Temporary structures	Where accommodation is used due to a lack of housing and is not the person’s usual place of residence
6	People living temporarily in conventional housing with family and friends due to lack of housing	Conventional housing, but not the person’s usual place of residence	Where accommodation is used due to a lack of housing and is not the person’s usual place of residence

Source: (FEANTSA, 2007[26]), ETHOS Light: European Typology of Homelessness and Housing Exclusion, www.feantsa.org/download/fea-002-18-update-ethos-light-0032417441788687419154.pdf.

There are, nevertheless, limits to the implementation of the ETHOS Light typology. For instance, some national homelessness statistics cover living situations that go beyond the ETHOS Light typology. For example, Australia’s statistical definition includes people living in an inadequate dwelling, a dwelling without tenure or with an initial tenure that is short and not extendable, or in a dwelling that does not allow them to have space for social relations. In New Zealand, the definition includes – in addition to people that could be considered in ETHOS Light 1, 2, 3, 5, and 6 – people living in uninhabitable housing, which is operationalised as people living in a dwelling that lacks one of six basic amenities: drinkable tap water, electricity, cooking facilities, a kitchen sink, bath or shower, and/or a toilet. Other types of living situations of relevance in some countries (such as allotments in Poland, or bed and breakfasts in the United Kingdom) are not explicitly included in the ETHOS Light typology.

A range of data collection methods to assess homelessness, but that may underreport or “miss” specific types of homelessness or socio-demographic groups

Official statistics are based on different data collection methods, including, among other things, street counts, survey-based methods, and administrative data (Box 2.2), and many countries rely on more than one approach to cover different types of living situations and socio-demographic groups. Each data collection approach presents strengths and weaknesses and is better suited to capture certain experiences of homelessness relative to others and generates varying levels of depth of information about individuals experiencing homelessness and their pathways into homelessness (cf. OECD Monitoring Framework (OECD, Forthcoming[6]) and corresponding Country Notes on Homelessness Data (OECD, 2024[27])). The data collection approaches discussed in this section are based on those that underpin the official homelessness statistics in OECD and EU countries, even if other collection approaches may be used.

Box 2.2. Six common approaches to collect data on homelessness in OECD and EU countries

The OECD Monitoring Framework (OECD, Forthcoming[6]) outlines six approaches to collect data on homelessness. These approaches are not mutually exclusive: in practice, countries frequently design surveys/counts that blend these approaches. These include:

• Street counts: an estimate of the number of people sleeping rough at a point-in-time;

• Service‑based methods: information obtained from a broad range of service providers that support people experiencing homelessness;

• Population censuses and Household surveys: a count or a sample of a given population at a point in time (e.g. Population Census; special module on homelessness in household survey);

• Administrative data: records collected by different institutions/organisations (e.g. health data, criminal justice records, social services data, etc.) and used to extrapolate the number of people experiencing homelessness;

• Advanced sampling methods: a statistical method, such as “capture‑recapture,” comparing independent samples from two or more sources of data to estimate the total number of people experiencing homelessness;

• By-name lists and Information management systems on homeless individuals: the collection of comprehensive demographic and identifying information on people experiencing homelessness, which may be collected via registry weeks.

The Framework summarises the main characteristics of each data collection approach, including, among other things, a general description of the approach; the type of count generated (point-in-time or flow); the source(s) of information (e.g. data from service providers; direct observation through a street count); the ETHOS Light groups typically covered (see Table 2.1); the strengths and limitations of the approach; the scope and depth of information collected; and common implementation challenges.

Nevertheless, the design and implementation of many standard data collection approaches often fail to account for some population groups, including, among others, women, youth/children, people who identify as LGBTI, racial and ethnic minorities, Indigenous Peoples, migrants, and people living in rural communities (see the OECD Monitoring Framework: Measuring Homelessness in OECD and EU countries (OECD, Forthcoming[6])). This can lead to an undercount of homelessness among such groups, and thus an underestimate of the total number of people experiencing homelessness. There are many reasons behind this:

• In some cases, the underreporting of certain socio-demographic groups is driven by the specific ways in which they experience homelessness, which are not well captured in standard data collection exercises. For instance, some socio-demographic groups are more likely to rely on informal networks of family and friends in a first instance, rather than formal emergency accommodation (e.g. shelters); this can be driven by the sentiment that traditional shelter settings and social services are not well adapted to their needs, do not feel safe, and/or because of past experiences or expectations of discrimination or stigma. Rural homelessness tends to be underreported because some of the most common data collection approaches are not well suited to assess homelessness in rural areas: street counts can be impractical in expansive, non-urban areas, and service‑based methods may not be effective or reliable in areas that are service‑poor.

• In other cases, the housing situation of socio-demographic groups that are characterised, for various reasons, by high mobility (such as Indigenous communities or Roma) may be harder to capture and quantify in official statistics.

• In addition, there may be limitations to the data collection approach: while service‑based counts are the most prevalent data collection approach among countries and offer many advantages, there is no systematic approach across countries to determining which types of services and emergency or temporary accommodation are included in data collection efforts, and which are left out, presenting a challenge for cross-country comparison.

Box 2.3. Challenges to measuring homelessness among migrants in OECD and EU countries

In OECD and EU countries, comprehensive, comparable data on homelessness among migrants do not exist. Fewer than half of OECD and EU countries report the share of migrants in national homelessness statistics. The other 20 countries do not report homelessness statistics disaggregated by origin. While most data on homelessness among migrants use citizenship as the basis of migrant status, some countries only include in official homelessness statistics migrants with legal residence, and/or professional and personal ties to the country.

Many of the broader methodological challenges that stymie homelessness measurement and cross-country comparisons (discussed throughout this Block) also affect the extent to which migrants are counted (or missed) in official homelessness statistics. These include differences in how countries define and measure homelessness, which likely result in an underestimate of homelessness generally, and often particularly so among migrants.

Yet in addition to these general challenges, there are different cross-country approaches to data collection that are specific to the case of migrants – notably relating to asylum seekers and refugees. In 11 OECD and EU countries, official homelessness statistics explicitly include people staying in temporary accommodation for asylum seekers and refugees; by contrast, in at least 20 countries, official homelessness statistics exclude individuals staying in such accommodation.

In countries for which disaggregated data on homelessness among migrants are available, estimates suggest that migrants are overrepresented among individuals experiencing homelessness. Data cannot, however, be readily compared across countries.

To strengthen the evidence on homelessness among migrants, governments may consider relying on multiple, co‑ordinated approaches to collect data on homelessness; expanding the types of surveyed accommodation and support services to include low-barrier services that are accessible to migrant populations; and, where feasible, including information on country of birth in homelessness data collection.

Note: For cross-country data availability on the share of migrants experiencing homelessness, refer to the detailed table in (OECD, 2024[28]).

Source: (OECD, 2024[28]), Challenges to measuring homelessness among migrants in OECD and EU countries, https://doi.org/10.1787/b9855842‑en.

Cross-country differences in the periodicity and geographic coverage of homelessness statistics

There are also important cross-country differences in the periodicity and the geographic coverage of official homelessness statistics. Some data collection approaches collect point-in-time (PIT) data (capturing a snapshot of homelessness in a specific location at a specific point-in-time), while others collect flow data (for instance, providing an estimate of the number of shelter users over the course of a year). PIT and flow data cannot be meaningfully compared. Moreover, the geographic coverage of official statistics also varies across countries and may only reflect the situation in the capital city (Iceland) or in selected cities and towns (such as Belgium, Colombia, France or Italy).

Common operational questions

Accordingly, there is considerable scope to improve the assessment of homelessness in OECD and EU countries, grounded in robust data and an understanding of the drivers of homelessness. Robust data should underpin all phases of policy making, including the development of homelessness strategies (Block 1) and monitoring and evaluation (Block 3); the implementation of effective policies relating to prevention (Block 4), long-term housing solutions (Block 5) and wraparound services (Block 6); as well as policy management, relating to funding and financing (Block 7), governance (Block 8) and the political economy of reform (Block 9).

The following set of operational questions is intended to guide policy makers and practitioners in identifying important focus areas to strengthen the assessment of homelessness in their country, city or community context; the relative importance of each component below will depend on the current state of data collection:

• What kind of statistics are needed to help address the policy objectives?

• What are key considerations for a statistical definition of homelessness?

• How to assess the main drivers of homelessness?

• How to improve the quality and coverage of homelessness data?

Recommendations about how – and why – to improve the communication of homelessness data to key stakeholders and the broader public are addressed in Block 9. These are complemented by guidance on some of the more technical dimensions of data collection, monitoring and reporting, and a self-assessment tool, in the OECD Monitoring Framework (OECD, Forthcoming[6]).

What kind of statistics are needed to help address the policy objectives?

Ensuring data collection is policy-relevant and identifying suitable data collection approaches for the country context

Data collection on homelessness should be designed and implemented with clear policy objectives in mind; in other words, data collection should serve a clear policy purpose, rather than be an end in itself. Such efforts can be enhanced by aligning data collection efforts with a clear, measurable policy commitment to end homelessness (or specific types of homelessness, such as rough sleeping or chronic homelessness). Identifying the policy objectives is a first step to determining what types of statistics are needed, and what types of data collection approaches are most suited and feasible.

The following questions can help determine the most appropriate data collection approaches to consider:

• What are the biggest policy priorities with respect to homelessness, and what form(s) of homelessness are to be addressed as a priority? A public policy objective to reduce rough sleeping could aim to assess people who are living rough, as well as people who may be at-risk of rough sleeping, and/or those who have recently experienced rough sleeping. To this end, different data collection approaches could be considered, ranging from street counts and service‑based methods, among others. In contrast, an effort to develop public policies to address “hidden homelessness” – generally understood to refer to people whose living situation corresponds to one of the categories outlined in the ETHOS Light typology (see Box 2.1), but who do not appear in official statistics on homelessness – would require a different approach to data collection (discussed further in the OECD Monitoring Framework: Measuring Homelessness in OECD and EU countries (OECD, Forthcoming[6])). Aligning data collection approaches with policy targets – as exemplified by the “Ending Homelessness Framework” of the Centre for Homelessness Impact in the United Kingdom, for instance – can be a useful approach.

• What type(s) of information and level of detail are needed for policy purposes? Collecting disaggregated data – including by type of living situation and key demographic characteristics – strengthens the evidence base about the scale and scope of the challenge, and can help to identify at-risk groups and be used to tailor interventions accordingly. Nevertheless, long questionnaires that aim to understand past and present living situations, drivers, services and housing needs and personal history can be a useful tool in some cases – but they are costly. The depth of information during the data collection process should be well aligned with policy needs and resources. Moreover, regardless of the approach(es) selected, privacy and confidentially must be assured, given the sensitive personal information that may be collected and its potential for misuse.

• What resources are available to undertake data collection efforts? This includes financial resources, as well as human resources, in terms of civil servants as well as non-public actors (NGOs, private sector, civil society, citizens) of who may, in some cases, contribute to design, implement and support data collection efforts (e.g. public authorities, research institute, NGO/private sector, civil society, etc.). Considerations should cover the data collection process, as well as the analysis and policy phases.

The following illustrations reflect the different types of data collection approaches that could be adopted; more information on countries’ data collection approaches are detailed in the OECD Country Notes on Homelessness Data (OECD, 2024[27]).

Using research and survey results to inform the development of a national homelessness strategy

Brazil’s National Survey on the Homeless Population, conducted in 2007 and 2008 and covering 71 cities and more than 300 000 individuals, helped to inform the development of a number of national policies and programmes, including the National Policy for the Homelessness Population. The policy, developed as a rights-based approach, notably included measures to ensure access to public health, education, social security, social assistance, housing, and other key areas; and the development of specialised help centres for people experiencing homelessness (World Without Poverty: Brazil Learning Initiative, 2016[29]).

Comprehensive surveys of people experiencing homelessness to facilitate an in-depth understanding of their needs and living conditions

In Colombia, the Census of Street Dwellers is a street count implemented over a five‑year period in different regions. The Census collects a wide range of information on people experiencing homelessness, including:

• Demographic information: sex, geographical location, place of birth, ethnicity, sexual orientation, level of education (literacy, highest level of educational attainment achieved).

• History of homelessness: location of initial period of homelessness, principal reason for homelessness, length of period of homelessness, principal reason for why they continue to experience homelessness.

• Survival and support networks: income generation, closest family member(s), receipt of aid, origin of aid, knowledge of social services and programmes.

• Health and substance use: human functioning, overall health, dental or health problems from the last 30 days, presence of diseases, consumption of psychoactive substances.

• Safety and security: overall sense of safety, aggressions of which they have been victims.

The extensive information aims to facilitate an in-depth understanding of the needs and living conditions of people experiencing homelessness, to help guide decision-making. Colombia’s dedicated strategy to prevent and combat homelessness, the Public Social Policy for Street Dwellers, leverages data from the Census of Street Dwellers on the socio-demographic characteristics of people experiencing homelessness to inform policy design.

Leveraging integrated academic knowledge to analyse statistical data

Health researchers in the United States sought to develop and validate a classification model of homelessness using a linked dataset of integrated administrative records from multiple state‑maintained databases which contained a sample of over 5 million individuals (Byrne et al., 2020[30]). The model used a logistic regression to predict cases of homelessness and identify targeted interventions to mitigate the risk of adverse health outcomes. This initiative underscores the potential value of collaborating with researchers across different sectors to help governments make better use of statistical data.

Using data and research to inform public policy

Across Belgium, data collection and research on homelessness, in co‑operation with local and regional governments, have enabled many local authorities to refine their homelessness policies by incorporating relevant new insights on rural vs. urban homelessness, gender, and (mental) health. These efforts are organised in collaboration with a diverse group of stakeholders, and the results are widely communicated, generating media coverage and informing public debate on policies aimed at addressing homelessness (Koning Boudewijnstichting, 2024[31]). Similarly, the city of Paris (France) developed new services targeting women experiencing homelessness following the first street count, Nuit de la solidarité, in 2018 (City of Paris (France), 2024[32]),

What are key considerations for a statistical definition of homelessness?

A clear, consistent statistical definition along ETHOS Light typology, where feasible, that aligns with data collection methods

Regardless of the scope of a country’s statistical definition of homelessness, the definition should i) provide a clear indication of the groups and the types of living situations that are included (drawing, for instance, on the different living situations proposed by the ETHOS Light typology), and ii) align with the data collection approaches en vigueur. In addition, data collection efforts that rely on surveys of support services should be clear about the scope of services to be surveyed; this could cover, among other things, emergency accommodation for people experiencing homelessness, food banks, day centres, shelters and support services for victims/survivors of domestic violence, accommodation for migrants, youth services, etc.

In Sweden, the statistical definition of homelessness includes people living in the following types of housing situations, which are in many cases relatively straightforward to map to ETHOS Light categories:

• Acute homelessness: rough sleeping; people staying in emergency accommodation, women’s shelters, etc. (ETHOS Light 1, 2).

• Longer-term housing solutions with unstable tenancy: people with transitional supported accommodated housing and a lease of a Housing First unit.

• Short-term housing solutions: people living with family or friends or people with temporary sublet contracts who have been in contact with social services (ETHOS Light 6).

• People in institutional care, category housing or penal institutions who are within three months of leaving and do not have a place to stay (ETHOS Light 4).

The definition includes clear requirements to be considered as experiencing homelessness (such as the three‑month period prior to release from an institution), as well as some level of specification on the types of institutions to be considered (such as category housing or penal institutions), helping to facilitate the standardisation of data collection efforts across service providers.

Official homelessness statistics in Australia are drawn from the Population Census, conducted every five years, and based on a statistical definition that is developed along six operational categories:

• People living in improvised dwellings, tents or sleeping out.

• People living in supported accommodation for the homeless.

• People staying temporarily with other households.

• People living in boarding houses.

• People in other temporary lodgings.

• People living in “severely” crowded dwellings (defined as living in a dwelling which requires five or more extra bedrooms to accommodate the people who usually live there).

Official homelessness statistics thus cover a wide array of living situations that would be categorised under ETHOS Light categories 1, 2, 3, 5, and 6, in addition to living situations that go beyond the ETHOS Light Typology (such as people living in severely crowded dwellings) and that are not considered in the homelessness definitions of most OECD or EU countries. In addition, Australia collects data through the Census on people who are “marginally housed” and may be at risk of homelessness, but nevertheless who are not classified as homeless and are thus not included in national homelessness statistics:

• People living in other crowded dwellings.

• People in other improvised dwellings.

• People marginally housed in caravan parks.

In Scotland (the United Kingdom), both the definition of homelessness, and therefore the statistics, have a prevention element, including people and/or households who will experience homelessness within the next two months (this could be extended to six months in forthcoming legislation). Statistics are produced every six months using Local Authority data, with a focus on:

• People and/or households who were sleeping rough prior to their application.

• People and/or households who are in temporary accommodation (and lengths of stay).

• People and/or households with children/dependents.

• Main reason for experiencing homelessness.

• Case characteristics, including support needs.

• Length of open cases from assessment to settled accommodation.

How to assess the main drivers of homelessness?

As discussed in the previous section, there are different pathways into homelessness, including structural factors, institutional and systemic failures, and individual circumstances (Box 2.1). These factors vary in salience across different country, city and community contexts. For instance, the high cost of living and a shortage of affordable and social may be a primary factor in contributing to housing instability and homelessness in one context, while gaps in social protection or in access to public support may be a more relevant factor in another context. Similarly, high rates of substance use in some countries and/or local contexts – including, notably, with respect to opioids (discussed in Block 6) – may be a significant contributing factor to (as well as a consequence of) homelessness, but play a relatively insignificant role in others. Further, systemic and institutional barriers to housing and social services are faced by some groups of the population. Understanding the distinct drivers of homelessness at the structural, systemic and individual levels allows policy makers and practitioners to improve their ability to prevent homelessness and to support people in crisis in developing sustainable pathways out of homelessness.

Including questions about individual pathways and trajectories in standard data collection efforts on homelessness

In Spain, the Survey of the Homeless People (EPSH) collects data on the socio-demographic characteristics of people using accommodation assistance centres and restoration centres collected through a questionnaire‑based interview, including the reasons why they began to experience homelessness and the time they have experienced homelessness for. In 2022, the primary reasons cited by interviewees for experiencing homelessness included starting over after migrating from another country, job loss and evictions. Approximately one‑third of respondents began experiencing homelessness less than a year prior to the survey, 27% had experienced homelessness for one and three years, and about 40% had experienced homelessness for over three years (INE, 2022[33]). Similarly, in Colombia, the Census of Street Dwellers incorporates a survey that seeks to determine the socio-demographic characteristics of people experiencing homelessness (OECD, 2024[34]), including the reasons for experiencing homelessness and its duration. In 2021, the two main reasons reported by respondents for experiencing homelessness were the use of psychoactive substances and family challenges and conflicts. Additionally, almost 62% of respondents had been experiencing homelessness for over five years (DANE, 2021[35]).

In Canada, the Co‑ordinated Point-in-Time Counts, Everyone Counts, gathers socio-demographic characteristics of people experiencing homelessness through a survey, including reasons for housing loss. In the Third Nationally Co‑ordinated Point-in-Time Counts, corresponding to 2020‑22, not having enough income for housing and substance use issues were the most commonly reported reasons. Regarding the duration of homelessness, 50% of respondents reported experiencing homelessness for the entire 12 months prior to the survey (Infrastructure Canada, 2024[36]).

Partnering with external researchers to conduct in-depth studies on the drivers of homelessness

In the United States, the Benioff Homelessness and Housing Initiative (BHHI) at the University of California assessed the drivers of homelessness in the California Statewide Study of People Experiencing Homelessness (CASPEH) (Benioff Homelessness and Housing Initiative, 2023[37]). In collaboration with local governments, the BHHI conducted the study at the request of the California Health and Human Services Agency. The study employed a mixed-methods approach that included surveys and in-depth interviews. It found that a combination of structural factors such as high housing costs and individual vulnerabilities (e.g. discrimination, family/social conflict, exposure to violence, incarceration) increase the risk of experiencing homelessness. The study highlights that, alongside violence, social conflict and household changes –whether violent or not –were significant precursors to homelessness, especially for those lacking affordable housing and adequate income. The study highlighted that the average age of people experiencing homelessness is increasing, marginalised groups tend to be overrepresented and there is a high prevalence of mental health conditions and substance use among people experiencing homelessness. The CASPEH survey provided important insights on the drivers of homelessness in California, highlighting the value of external expertise in homelessness research.

How to improve the quality and coverage of homelessness data?

Relying on multiple data collection approaches to assess different forms of homelessness

There is no single experience of homelessness, nor a typical profile of a person experiencing homelessness. Research has found that people may experience homelessness on a chronic or temporary basis; their living situations are diverse, ranging from sleeping rough, to staying in shelters or temporary accommodation, to sleeping with family and friends; and that different socio-demographic groups (such as women, youth, LGBTI and migrants) may be more or less likely to experience different types of homelessness relative to others. As mentioned, women are less likely to sleep rough, and more likely to first stay with family or friends before turning to emergency accommodation; others, including LGBTI or migrants, may face barriers and/or discrimination in accessing shelters and services for the homeless. Many annual point-in-time counts tend to take place in late autumn and winter, which risks underestimating the number of people sleeping rough, when weather conditions are less temperate. To reflect this diversity and overcome the data limitations, many countries thus rely on multiple data collection approaches to develop official homelessness statistics.

For instance, the United States assesses different forms of homelessness through five data collection approaches, which each covers different subgroups and ETHOS Light categories:

• The Annual Point-in-Time Count, which collects information from both sheltered and unsheltered individuals across the United States at a single point in time using a street and shelter count performed by volunteers.

• The Annual Homeless Assessment Report to Congress, which presents information from the Homeless Management Information System (HMIS), a case‑level information system managed by each Continuum of Care (CoC) on clients to homelessness services.

• The US Census and American Community Survey, which includes outreach efforts to enumerate people experiencing homelessness through a questionnaire.

• An annual count on the number of adults and children seeking services from domestic violence shelter programmes, which is conducted through a single 24‑hour survey period.

• In addition, the US Department of Education’s National Center for Education monitors the number of homeless students identified by public schools each year. Their definition includes any child who lacks a fixed, regular, and adequate nighttime residence (Lowell and Hanratty, 2022[38]; U.S. Department of Education, 2023[39]).

In Czechia, a comprehensive census conducted in 2022 built on a previous census and expanded the coverage of ETHOS categories (OECD, 2024[40]):

• Conducted in 2019, the Homeless People Census sought to support the creation of a database with socio-demographic information on people experiencing homelessness, with a specific focus on those without a roof and without an apartment – ETHOS Light 1, 2, 3 and 4 -. For this, street counts were conducted, and service‑based methods, capture‑recapture methods were utilised.

• The Census of People from Selected Categories of the ETHOS Classification builds on the Homeless People Census by focusing on groups that are harder to detect, such as people staying in unconventional dwellings and people staying with family and friends (i.e. ETHOS Light 5 and 6) through administrative data and service‑based methods.

Further, combining data collection approaches that provide both point-in-time and flow estimates allows for a better understanding of the transitional nature of homelessness. In Canada, six collection methods are used to collect information on people experiencing homelessness:

• Everyone Counts, a co‑ordinated point-in-time count of rough sleepers, people staying in shelters, and people in transitional housing programmes.

• National Shelter Study, an annual estimate of the number of people who access emergency shelters over the course of the year.

• Shelter Capacity Report, a report on the capacity of emergency homeless shelters, transitional housing programmes, and domestic violence shelters.

• Canadian Housing Survey, a survey on housing needs and households’ experiences, including previous experiences of homelessness.

• National Census, which provides point-in-time information from people staying in shelters.

• Homeless Individuals and Families Information System (HIFIS), a web-enabled case management system for service providers to collect information on individuals experiencing homelessness. The data from this system (and similar systems) contribute to the National Shelter Study and the Shelter Capacity Report.

The combination of different approaches has also helped to cover a broad range of information on potentially hard-to-reach subgroups of people experiencing homelessness (e.g. unaccompanied youth, people staying with friends or family, LGBTI individuals, and First Nations, Métis or Inuit people). Nonetheless, coverage limitations exist across subgroups.

Casting a wider net in surveyed service providers to enumerate people who are at risk of, or experiencing, homelessness and harder-to-reach socio-demographic groups

Surveying a wide range of services in service‑based methods can help capture a broader spectrum of homelessness experiences and improve coverage of individuals who are often underreported in official statistics. This includes people who may seek to avoid detection during street counts by selecting more concealed locations, as well as those who access services that are not covered by existing approaches (e.g. service‑based methods). Conducting preliminary qualitative research (including through semi-structured interviews, for instance) to identify locations and services where specific, hard-to-reach populations are more likely to be found can facilitate more effective sampling. Specific adaptations to the design and implementation of street counts and service‑based methods can also be envisaged.

In Germany, national statistics on people experiencing homelessness are based on two complementary survey-based collection methods, which have been carefully designed to minimise the risk of double counting:

• The Annual Reporting of People Sleeping in Shelters for the Homeless, which provides annual point-in-time information on people staying in registered shelters.

• The Biennial Survey to Count People Sleeping Rough and People Staying with Family/Friends, which collects information from a broad range of service providers (food banks, health clinics, youth centres, etc.) that aim to support people sleeping rough and people staying with family or friends.

In Belgium, a wide array of services is involved in the data collection efforts, including general social services not directly linked to homelessness. Data collection involves night shelters, public and non-profit social services, health institutions like hospitals, psychiatric facilities and refugee centres. Further, low-threshold services, including drop-in centres and social restaurants and social housing companies play a key role in data collection on homelessness, ensuring people experiencing hidden forms of homelessness are counted.

The OECD Monitoring Framework: Measuring Homelessness in OECD and EU Countries (OECD, Forthcoming[6]) discusses a range of additional strategies that can improve data coverage, notably of hard-to reach groups, and to assess “hidden homelessness”, including:

• Modifying existing data collection approaches to better account for the homelessness experiences and service usage patterns of hard-to-reach groups. For instance, in London (the United Kingdom), a collaboration among NGOs and local authorities led to the Women’s Census of Rough Sleeping, designed to better capture the extent of rough sleeping among women, since women are often less visible in street counts and less likely to engage with outreach teams. Outreach services identified locations frequented by women experiencing homelessness, and conducted the count over a week, rather than a single night, between 7 PM and 7 AM, including at least one daytime shift. Outreach workers were also trained in “trauma‑informed outreach” to enhance the sense of safety and control among the women they assisted.

• Adjusting the frequency of data collection (where relevant). Resources-permitting, it can also be relevant to consider collecting data on a more regular basis, in cases where data collection is infrequently or irregularly conducted. For instance, counts that do not occur on at least an annual or bi‑annual basis could be exercised more regularly. Further, even annual point-in-time counts have their limitations: conducting point-in-time counts (such as street counts) biannually, for instance both a winter and summer count, can help account for changing environmental and social conditions.

• Linking administrative data across systems, such as social services, housing authorities, and healthcare providers, can help identify hard-to-reach populations, including individuals who may not engage with traditional homelessness services. Administrative data from multiple agencies can provide information about housing status over time and reveal patterns of service usage, helping to detect individuals facing housing insecurity who might otherwise be overlooked. For example, in Finland, the Housing Finance and Development Centre (ARA) circulates comprehensive surveys to municipalities, which collect and report data from various sources, including social welfare and housing service registers, municipal rental housing applicant registers, the Social Insurance Institution’s register, and the Digital and Population Data Services Agency’s Population Information Register. This broad range of data sources enables the collection of nationwide homelessness data, covering five categories of the ETHOS Light framework, including ETHOS Light 6 (OECD, 2024[27]).

Fundamentals for success

Robust homelessness data are the foundation to understanding the scope and scale of the challenge, and to developing effective public policies. To address the persistent methodological challenges and data gaps in the field of homelessness, there are a number of recommendations for policy makers and practitioners to strengthen homelessness data collection efforts. This Block provides guidance and good practice examples to improve data quality, consistency and comparability. Recommendations thus relate to the identification of different data collection approach(es), the rationale for selecting them and the coherence across different approaches; ways to assess the main drivers of homelessness; considerations for developing a robust statistical definition; as well as decisions about the type(s) of information, and level of detail, that is needed for policy purposes. Across OECD and EU countries, many governments are rethinking their approaches to collect homelessness data, proposing innovative methods, and investing in more robust data collection efforts.

Building on the operational issues and good practice illustrations described above, the following recommendations can thus help policy makers and practitioners strengthen the assessment of homelessness:

• Ensure that homelessness data collection is policy-relevant – that is, that data collection is designed and implemented to meet a clear policy purpose. In particular, aligning data collection efforts with a clear, measurable policy commitment to end homelessness (or specific types of homelessness) can be helpful.

• Develop a clear, consistent statistical definition of homelessness, upon which data collection efforts are based, drawing on the ETHOS Light Typology where feasible.

• Collect disaggregated data by different types of homelessness (e.g. ETHOS typology) and key demographic characteristics of relevance in a given country context to facilitate in-depth assessments and to tailor interventions accordingly.

• Undertake efforts (including through partnerships with competent research entities and NGOs) to assess the structural, systemic, institutional and/or individual drivers of homelessness in your country, city or community context, to improve the capacity to prevent homelessness and help people exit homelessness.

• Establish a standardised, consistent data collection and monitoring system, which may draw on multiple data collection approaches, and improve data coverage of hard-to-reach groups (including by collecting data from a broad range of service providers, e.g. beyond emergency shelters and temporary accommodation for people experiencing homelessness).

Further discussion of the additional means to strengthen data collection, reporting, and monitoring, along with a self-assessment tool to identify opportunities to strengthen data collection, is detailed in the OECD Monitoring Framework: Measuring Homelessness in OECD and EU countries (OECD, Forthcoming[6]) and in Block 9:

• Explore opportunities to measure “hidden homelessness” and to improve data coverage of hard-to-reach groups in homelessness statistics.

• Ensure the privacy and confidentiality of individuals experiencing homelessness by obtaining informed consent, storing data securely, and using anonymised data where suitable.

• Regularly report data on people experiencing homelessness and make key indicators publicly available to facilitate research and policy development and promote transparency and accountability (Block 9).

References

[11] Baptista, I. and E. Marlier (2019), Fighting homelessness and housing exclusion in Europe: A study of national policies, https://doi.org/10.2767/624509.

[37] Benioff Homelessness and Housing Initiative (2023), The California Statewide Study of People Experiencing Homelessness, UCSF.

[16] Bonakdar, A. et al. (2023), “Child protection services and youth experiencing homelessness: Findings of the 2019 national youth homelessness survey in Canada”, Children and Youth Services Review, Vol. 153, p. 107088, https://doi.org/10.1016/j.childyouth.2023.107088.

[19] Bretherton, J. (2017), “Reconsidering Gender in Homelessness”, European Journal of Homelessness, Vol. 11, https://www.feantsaresearch.org/download/feantsa-ejh-11-1_a1-v045913941269604492255.pdf.

[1] Busch-Geertsema, V., D. Culhane and S. Fitzpatrick (2016), “Developing a global framework for conceptualising and measuring homelessness”, Habitat International, Vol. 55, pp. 124-132, https://doi.org/10.1016/j.habitatint.2016.03.004.

[32] City of Paris (France) (2024), Quelles avancées depuis la première Nuit de la Solidarité ?, https://www.paris.fr/pages/quelles-avancees-depuis-la-premiere-nuit-de-la-solidarite-22875.

[13] Colburn, G. and C. Page Aldern (2022), Homelessness Is a Housing Problem, University of California Press, https://www.ucpress.edu/books/homelessness-is-a-housing-problem/.

[35] DANE (2021), Censo Habitantes de la Calle, https://www.dane.gov.co/files/investigaciones/boletines/censo-habitantes-calle/caracterizacion-CHC-2021.pdf.

[4] Drilling, M. et al. (2020), Measuring Homelessness by City Counts – Experiences from European Cities, https://www.feantsaresearch.org/public/user/Observatory/2021/EJH_14-3/EJH_14-3_A4_web2.pdf.

[20] Ecker, J., T. Aubry and J. Sylvestre (2019), “Pathways Into Homelessness Among LGBTQ2S Adults”, Journal of Homosexuality, Vol. 67/11, pp. 1625-1643, https://doi.org/10.1080/00918369.2019.1600902.

[14] FAP (2019), L’état du mal-logement en France 2019 : Rapport annuel #24, https://www.fondation-abbe-pierre.fr/documents/pdf/rapport_complet_etat_du_mal_logement_2019_def_web.pdf (accessed on 14 May 2019).

[26] FEANTSA (2007), ETHOS Light: European Typology of Homelessness and Housing Exclusion, https://www.feantsa.org/download/fea-002-18-update-ethos-light-0032417441788687419154.pdf (accessed on 24 July 2024).

[15] Gaetz, S. et al. (2016), Without a Home: The National Youth Homelessness Survey, Canadian Observatory on Homelessness Press, https://homelesshub.ca/sites/default/files/WithoutAHome-final.pdf.

[2] Hermans, K. (2023), Towards a harmonised homelessness data collection and monitoring strategy at the EU-level. Position paper prepared for the European Mutual Learning Event of the European Platform on Combatting Homelessness, https://event5.homeless-platform-events.eu/media/fphfbq4u/hermans_towards-a-harmonised-homelessness-data-collection.pdf.

[33] INE (2022), Encuesta a las personas sin hogar. Año 2022, https://www.ine.es/prensa/epsh_2022.pdf.

[36] Infrastructure Canada (2024), veryone Counts 2020-2022 – Results from the Third Nationally Coordinated Point-in-Time Counts of Homelessness in Canada, https://housing-infrastructure.canada.ca/homelessness-sans-abri/reports-rapports/pit-counts-dp-2020-2022-results-resultats-eng.html.

[21] Johnson, G. et al. (2015), Examining the relationship between structural factors, individual characteristics, and homelessness, http://www.ahuri.edu.au/publications/projects/p53042 (accessed on 18 December 2019).

[31] Koning Boudewijnstichting (2024), Telling dak- en thuisloosheid. Globaal rapport 2023, https://kbs-frb.be/nl/telling-dak-en-thuisloosheid-globaal-rapport-2023.

[8] Lee, B., M. Shinn and D. Culhane (2021), “Homelessness as a Moving Target”, The ANNALS of the American Academy of Political and Social Science, Vol. 693/1, pp. 8-26, https://doi.org/10.1177/0002716221997038.

[38] Lowell, W. and M. Hanratty (2022), “Who Counts? Educational Disadvantage among Children Identified as Homeless and Implications for the Systems That Serve Them”, Social Service Review, Vol. 96/4, pp. 581-616, https://doi.org/10.1086/722003.

[22] Ministry of Housing, C. (2019), Causes of Homelessness and Rough Sleeping: Rapid Evidence Assessment.

[7] OECD (2024), Affordable Housing Database, OECD, Paris, https://www.oecd.org/en/data/datasets/oecd-affordable-housing-database.html.

[28] OECD (2024), Challenges to measuring homelessness among migrants in OECD and EU countries, OECD Publishing, Paris, https://doi.org/10.1787/b9855842-en.

[34] OECD (2024), Country note: Data on homelessness in Colombia, OECD, Paris, https://webfs.oecd.org/Els-com/Affordable_Housing_Database/Country%20notes/Homelessness-COL.pdf.

[40] OECD (2024), Country note: Data on homelessness in Czechia, OECD, Paris, https://webfs.oecd.org/Els-com/Affordable_Housing_Database/Country%20notes/Homelessness-CZE.pdf.

[5] OECD (2024), HC3.1. Population experiencing homelessness, OECD, Paris, https://webfs.oecd.org/Els-com/Affordable_Housing_Database/HC3-1-Population-experiencing-homelessness.pdf.

[27] OECD (2024), OECD Country Notes on Homelessness Data, OECD, Paris, https://www.oecd.org/en/topics/sub-issues/affordable-housing/homelessness.html.

[24] OECD (2023), Supporting Lives Free from Intimate Partner Violence: Towards Better Integration of Services for Victims/Survivors, OECD Publishing, Paris, https://doi.org/10.1787/d61633e7-en.

[10] OECD (2020), Better data and policies to fight homelessness in the OECD. Policy Brief on Affordable Housing, OECD Publishing, Paris, https://www.oecd.org/en/publications/better-data-and-policies-to-fight-homelessness-in-the-oecd_0eef075a-en.html (accessed on 16 March 2020).

[9] OECD (2015), Integrating Social Services for Vulnerable Groups: Bridging Sectors for Better Service Delivery, OECD Publishing, Paris, https://doi.org/10.1787/9789264233775-en.

[6] OECD (Forthcoming), OECD Monitoring Framework: Measuring Homelessness in OECD and EU Countries, OECD, Paris.

[17] Olivet, J. et al. (2021), “Racial Inequity and Homelessness: Findings from the SPARC Study”, The ANNALS of the American Academy of Political and Social Science, Vol. 693/1, pp. 82-100, https://doi.org/10.1177/0002716221991040.

[18] Paul, D. et al. (2019), “Racial discrimination in the life course of older adults experiencing homelessness: results from the HOPE HOME study”, Journal of Social Distress and Homelessness, Vol. 29/2, pp. 184-193, https://doi.org/10.1080/10530789.2019.1702248.

[23] Piat, M. et al. (2015), “Pathways into homelessness: Understanding how both individual and structural factors contribute to and sustain homelessness in Canada”, Urban Studies, Vol. 52/13, pp. 2366-2382, https://doi.org/10.1177/0042098014548138.

[12] Quigley, J., S. Raphael and E. Smolensky (2002), Homeless in America, Homeless in California, https://escholarship.org/uc/item/4v61c0ws (accessed on 15 May 2019).

[30] Sartorius, B. (ed.) (2020), “A classification model of homelessness using integrated administrative data: Implications for targeting interventions to improve the housing status, health and well-being of a highly vulnerable population”, PLOS ONE, Vol. 15/8, p. e0237905, https://doi.org/10.1371/journal.pone.0237905.

[25] Sullivan, C. et al. (2023), “Domestic Violence Housing First Model and Association With Survivors’ Housing Stability, Safety, and Well-being Over 2 Years”, JAMA Network Open, Vol. 6/6, p. e2320213, https://doi.org/10.1001/jamanetworkopen.2023.20213.

[39] U.S. Department of Education (2023), Identifying and Supporting Students Experiencing Homelessness from Pre-School to Post-Secondary Ages, U.S. Department of Education.

[3] van Ooijen, C., B. Ubaldi and B. Welby (2019), “A data-driven public sector: Enabling the strategic use of data for productive, inclusive and trustworthy governance”, OECD Working Papers on Public Governance, No. 33, OECD Publishing, Paris, https://doi.org/10.1787/09ab162c-en.

[29] World Without Poverty: Brazil Learning Initiative (2016), The National Survey on the Homeless Population, https://wwp.org.br/wp-content/uploads/2016/11/The-National-Survey-on-the-Homeless-Population-WWP-Series-Use-Reports.pdf.