This chapter examines data collection approaches in selected OECD countries to measure unsheltered homelessness and temporary stays with family or friends. Drawing on experiences from Australia, Canada, Czechia, Denmark, Finland, Germany and Scotland (United Kingdom), it reviews the use of street counts, service-based methods, administrative data, population censuses, household surveys and information management systems, and explores their practical implementation. In each case, the chapter summarises the institutional context and objectives, describes in detail the data collection approach and its implementation, highlights the main strengths and limitations, and concludes with a set of overarching lessons to consider for each method.
Measuring Unsheltered and Hidden Homelessness in Germany
3. Approaches to measure unsheltered and hidden homelessness: A review of selected international experience
Copy link to 3. Approaches to measure unsheltered and hidden homelessness: A review of selected international experienceAbstract
Street counts as a basis for measuring unsheltered homelessness
Copy link to Street counts as a basis for measuring unsheltered homelessnessStreet counts are the most frequently used approach to measure ETHOS Light 1 in OECD countries (Table 2.3) and have become more widely used in recent years, particularly in European cities (Teller et al., 2025[1]; OECD, 2025[2]). As outlined in the OECD Monitoring Framework to Measure Homelessness (OECD, 2025[2]), street counts provide a snapshot of the extent of unsheltered homelessness at a specific point-in-time by enumerating individuals sleeping in public spaces and/or on the streets, typically on a single night. They are primarily designed to assess the extent of ETHOS Light 1. Depending on definitions and implementation, they may also include, to some extent, some living situations covered by ETHOS Light 5 (i.e. people living in non-conventional dwellings, such as caravans, cars or tents). Street counts are often conducted at the local level, typically as city-led initiatives, which may be part of a broader national measurement framework.
Street counts offer several advantages (OECD, 2025[2]). Depending on their design, they can provide a relatively low-cost entry point into homelessness measurement and, when repeated consistently over time, can generate geographically specific estimates of unsheltered homelessness that facilitate monitoring of trends over time. They can also reach individuals who are not necessarily in touch with homelessness support services – a group that is often entirely missed by other measurement approaches. In addition, the frequent involvement of volunteers (although this is not always the case) can contribute to raising public awareness and reducing potential stigma associated with homelessness, though appropriate training is required. Street counts can also play a complementary role within broader data systems. In contexts where service-based methods may miss individuals who do not engage with services, street counts can help identify gaps in service coverage (Teller et al., 2025[1]). Similarly, in countries relying primarily on population censuses for national homelessness estimates, outreach-based enumeration approaches where teams support individuals in completing surveys can serve as a hybrid model, combining elements of street counts and census-based data collection (Section 3.4).
However, important limitations must also be acknowledged (OECD, 2025[2]). Street counts are inherently constrained by their point-in-time design which makes them less likely to capture short-term or episodic homelessness. They also do not account for the mobile and hidden nature of homelessness, as unsheltered individuals may be in places that are inaccessible or not visible to enumerators. While there are practices in place to improve coverage, this may still lead to systematic undercounting. Consequently, street counts should always be interpreted as a snapshot providing lower-bound estimates of homelessness (Teller et al., 2025[1]). Further, certain population groups are particularly likely to be undercounted (Section 2.1.1), including women who may avoid visible locations for safety reasons, undocumented migrants who may avoid contact with official-looking enumerators, LGBTI individuals who may fear discrimination or violence, and more mobile populations. Further, street counts offer limited depth of information on the profile of people experiencing homelessness where only headcounts are conducted (OECD, 2025[2]; Hermans et al., 2025[3]). Street counts are generally better suited to urban environments, while implementation in rural areas remains difficult (OECD, 2025[2]).
The design of street counts can range from basic headcounts that aim to provide an overall estimate of the scale of homelessness, to more elaborate approaches that incorporate surveys to collect information on the characteristics and needs of the population (Teller et al., 2025[1]). While survey-based approaches partially address the limitation of limited profile data, they require informed consent and may reduce participation rates (Hermans et al., 2025[3]).
As highlighted by the European Homelessness Counts pilot project (Box 2.1), which contributed to standardising street count methodologies across European cities, the implementation of street counts raises numerous operational issues that affect the results. For example, street counts require the delineation of geographic zones to identify the exact areas to be covered. The timing of the count, including the time of year and day, as well as the length of the exercise, considerably affect the results. The reliability of the count also depends on the familiarity and preparation of the enumerators, who must be recruited and trained. There are also ethical considerations, such as securing voluntary participation of people experiencing unsheltered homelessness and respecting their privacy during data collection (Teller et al., 2025[1]). These implementation challenges contribute to criticism of the method in some local and national contexts (Hermans et al., 2025[3]).
Canada: Everyone Counts Point-in-Time Count
National homelessness estimates in Canada are based on the Everyone Counts Point-in-Time (PiT) Count, a nationally co‑ordinated effort that combines an annual enumeration (street count) with a triennial survey of people experiencing homelessness (OECD, 2026[4]). The most annual recent enumeration was conducted in fall 2025, marking the fifth nationally co‑ordinated edition, with 75 communities participating in all provinces and territories (Housing, Infrastructure and Communities Canada, 2025[5]).
Institutional context and objectives
The PiT Count, introduced in 2016, has become a key component of Canada’s homelessness data collection system, complementing the National Shelter Study that continues to offer valuable longitudinal data (Housing, Infrastructure and Communities Canada, 2025[5]). While there is no legal obligation at the national level to collect data on homelessness, participation in Everyone Counts is required for most communities receiving federal funding under Reaching Home (OECD, 2026[4]).
The PiT Count is led at the community level by the designated local organisations funded by Canada’s national homelessness strategy; these include Reaching Home Community Entity (CE), the Indigenous Homelessness CE, a collaboration between the two, or a contracted third party (Housing, Infrastructure and Communities Canada, 2025[5]). Housing, Infrastructure and Communities Canada (HICC) serves as the federal co‑ordinating body, setting the methodology, providing guidance and templates, receiving community-level data, and publishing national results.
Everyone Counts serves two complementary functions. First, the annual enumeration provides a regular assessment of the impact of policy, programme, and economic changes across jurisdictions by estimating the number of people experiencing homelessness in a defined geographical area on a single night; it is used as a tool to measure the progress towards ending homelessness (Housing, Infrastructure and Communities Canada, 2025[5]). Second, the triennial Survey on Homelessness collects more detailed information on the characteristics and experiences of people affected by homelessness to help the government and community organisations better understand and serve individuals experiencing homelessness in Canada (Housing, Infrastructure and Communities, 2024[6]).
Data collection approach and methodology
The PiT count enumerates people experiencing homelessness in shelters, encampments, other unsheltered locations, and transitional housing on a single night. The count covers ETHOS Light 1, 2, 3 and 5. Some communities additionally capture people experiencing homelessness in institutional settings (ETHOS Light 4), and people experiencing hidden homelessness,1 corresponding to ETHOS Light 6. However, due to the absence of a consistently established methodology as well as inconsistent access to data across jurisdictions, ETHOS Light 4 and 6 do not feed into the national estimates (Housing, Infrastructure and Communities Canada, 2025[5]).
The PiT count is governed by a set of core standards for participation to ensure methodological consistency across communities. A local PiT Co‑ordinator – often a municipal employee – is responsible for the overall operational management of the count, setting the timeframe and geographical scope, training staff, and ensuring the quality of the data collected for official reporting to the government. In addition, a local PiT Count Committee is established, comprised of diverse community stakeholders, to support the co‑ordinator and promote acceptance within the community. The enumeration takes place on a single night between 1 October and 30 November.
The methodology for the unsheltered data collection (ETHOS Light 1) can take different forms, depending on community size, geography and resources:
full coverage of the entire area or specific subareas;
a known-locations approach, canvassing areas known to be frequented by people experiencing homelessness;
a statistical sampling approach that is typically used in larger communities.
In communities that feature urban, suburban and/or rural areas, a mixed approach can be used where appropriate. This is the case, for instance, in the Toronto metropolitan area, which relies on a full-coverage survey of the downtown area and a statistical sample of suburban areas.
While approaches can vary across communities, the enumeration is typically carried out by teams of two to three people, drawn from volunteers of local community partners or the general public, ideally including at least one person with experience working with people experiencing homelessness. People experiencing homelessness are identified primarily through a short set of core screening questions, which establish whether they have already been approached by another team to prevent double counting, their willingness to participate, and their overnight location. Where it is not possible to engage with an individual, observed homelessness may be used as a supplementary method, but only as a last resort, as it risks producing less reliable data (Housing, Infrastructure and Communities Canada, 2025[5]).
In survey years (every three years), a survey is administered in addition to the enumeration. The survey consists of 15 core questions covering sociodemographic characteristics and experiences related to housing and homelessness. The survey can be conducted over a period of up to one month following enumeration night, with the option given to local communities to add own questions (Housing, Infrastructure and Communities Canada, 2025[5]).
Strengths and limitations
The Canadian PiT framework combines consistent national standards, along with clear implementation guidance with some degree of flexibility to allow for local adaptation, thus facilitating comparable snapshots of homelessness over time and enabling participation across diverse communities. In this context, the policy objective of measuring progress toward ending homelessness is clearly defined. At the same time, there is transparency about the inherent limitations of PiT methodology: the enumeration is explicitly positioned as a tool to generate a minimum estimate and a one‑night snapshot, rather than a comprehensive measure of all people experiencing homelessness over time (Housing, Infrastructure and Communities Canada, 2024), in line with general limitations of street counts (Teller et al., 2025[1]; OECD, 2025[2]). For this reason, Canada relies on multiple data sources to provide a more comprehensive picture, including national shelter use statistics and monthly community system flow data. Canada’s PiT framework has also evolved over time to reflect changing realities of how people experience homelessness: for instance, encampments, which are defined as outdoor locations with a group of visible structures where two or more individuals are staying, have been introduced as a distinct sub-category (Housing, Infrastructure and Communities Canada, 2024).
While the degree of flexibility afforded to communities to operationalise the count allows for participation across diverse communities, it also means that results are not fully comparable across, as well as within, communities. More specifically, the different methodological options to determine the geographic parameters of the count – which range from full coverage of the community to selected locations to sampling approaches – can significantly influence results. For example, while the most recent results (2025) suggested little change from the results from the prior count (2023-2025), actual increases may be partially masked, as approximately a quarter of communities reported reduced geographic coverage of unsheltered areas relative to their previous count (Government of Canada, 2026[7]).
Given that street counts are more likely to be implemented at the local level in cities, the Canadian PiT enumeration is an interesting case, demonstrating the potential of street counts to be scaled at the national level in a large, federal country. For its most recent PiT enumeration, held in 2026, Canada expanded its support to more rural and remote communities to widen the PiT Count’s coverage of the Canadian population. In the Canadian case, the conditions for effective implementation include a high level of ownership and flexibility, alongside a core set of standards given to the communities, as well as a clear understanding that this approach only generates a snapshot of homelessness at a specific point in time, rather than a precise, comprehensive estimate of homelessness in Canada.
Lessons from street counts
Street counts are the most direct way of measuring unsheltered homelessness. However, they can only count people experiencing homelessness who are visible to enumerators at the time the count is conducted, which is often a single night. Therefore, they are likely to underestimate the true scale of unsheltered homelessness – specifically short-term unsheltered homelessness – and, by their design, do not provide insights into most other forms of homelessness (including ETHOS Light 6).
Since street counts are generally designed for implementation at local level, scaling up to national level, as demonstrated by the Canadian PiT Count, requires common core standards and strong co‑ordination, as well as clear local ownership and a certain degree of flexibility for local implementation. Specific community engagement strategies and the deployment of dedicated staff can be used to reach population groups at a higher risk of being undercounted, which is also demonstrated by census field operations to measure ETHOS Light 1 (see Section 3.4).
service-based methods as a basis for measuring unsheltered homelessness and temporary stays with family or friends
Copy link to service-based methods as a basis for measuring unsheltered homelessness and temporary stays with family or friendsservice-based methods are the most common approach to collect data on all forms of homelessness and housing exclusion in OECD and EU countries (Section 2.1) and are frequently used to assess the extent of unsheltered homelessness (ETHOS Light 1) and temporary stays with family or friends (ETHOS Light 6). They rely on information obtained from a more or less broad range of service providers that support people experiencing, or at risk of, homelessness.
In general, service-based methods can generate in-depth information on the profiles and needs of individuals experiencing homelessness, as well as an estimate of the flows of people experiencing homelessness over a given period. Depending on their design and implementation, they can be tailored to encompass any of the ETHOS Light categories. Specifically relevant to this report, the selection of the types of services that form the basis of this method can be effective in assessing homelessness among harder-to-reach populations, including those who are less likely to sleep rough or stay in emergency shelters, and more likely to stay with family or friends (OECD, 2025[2]).
However, a key limitation is that service-based methods do not account for people who are not (for whatever reason) in contact with services, and they are less effective or reliable in service‑poor areas. In this way, reliance on service-based methods can underestimate homelessness among some socio-demographic groups that do not have access to, do not feel comfortable accessing, or face discrimination accessing services for people experiencing homelessness. Further, there is no systematic approach to determining which types of services and accommodation are included in data collection efforts. This can limit comparability of measurement results (OECD, 2025[2]).
Germany: A biennial survey with random sampling
The headline estimate of homelessness in Germany is derived from a mixed approach, which combines an annual count of all people staying in nationally registered shelters and a biennial survey to enumerate people sleeping rough and people staying with family or friends. Both data collection approaches can be classified as service-based point-in-time counts (OECD, 2026[4]).
Institutional context and objectives
The 2020 Homelessness Reporting Act (Wohnungslosenberichterstattungsgesetz) provides the legal framework for collecting homelessness data. It mandates the collection of data on sheltered homelessness at the federal level on an annual basis (ETHOS Light 2 and 3), as well as supplementary reporting on unsheltered homelessness (ETHOS Light 1) and people staying with family or friends (ETHOS Light 6) on a biennial basis. It also requires that the government conduct a feasibility study to assess the potential expansion of homelessness data collection as part of the first supplementary report, and a further assessment as part of the third supplementary report (Section 2.2).
The annual point-in-time count of people in homelessness shelters is led by the Federal Statistical Office (Destatis). Nationally registered shelters report the number of shelter users on 31 January into a centralised system, generating national estimates for ETHOS Light 2 and 3 (OECD, 2026[4]). According to the Homelessness Reporting Act, the objective of this annual statistic is to improve the federal government’s reporting on poverty and wealth and to strengthen the evidence base for political action.
The biennial survey to enumerate people sleeping rough (ETHOS Light 1) and people staying with family or friends (ETHOS Light 6) is commissioned to a research institute. The first and second editions (2022, 2024) were carried out by the research consortiums GISS/Kantar and GISS/Verian respectively, with the third edition (2026) commissioned to the German Institute for Urban Affairs (DIFU). For the 2022 report, the responsibility was with the Federal Ministry of Labour and Social Affairs (BMAS) but since responsibility for the homelessness policy field transferred to the Federal Ministry for Housing, Urban Development and Building (BMWSB), the 2024 report onward has been commissioned by the BMWSB (Bundesministerium für Arbeit und Soziales, 2022[8]; Bundesministerium für Wohnen, Stadtentwicklung und Bauwesen, 2024[9]).
Per the Homelessness Reporting Act, the purpose of the biennial reporting is to generate information and analysis on forms of homelessness that fall outside the scope of the annual statistic, including information about the causes and duration of homelessness, contact with the support system, experiences of violence, health status, and specific focus topics that may vary by year (e.g. the situation of homeless migrants in 2024) (Bundesministerium für Wohnen, Stadtentwicklung und Bauwesen, 2024[9]).
Data collection approach and methodology
The biennial supplementary reporting covers two ETHOS Light categories, defined as follows:
ETHOS Light 1: people who, at the time of interview or in at least one of the seven preceding nights, slept on the street or in makeshift accommodation. This also includes longer-term or permanent living in tents, cars, derelict buildings, or garages.2
ETHOS Light 6: people who, in at least one of the seven nights before the survey, (i) were not institutionally accommodated, (ii) did not qualify as ETHOS Light 1, and (iii) due to hardship had no dwelling of their own secured by rental contract or ownership and had instead sought temporary refuge with friends or family without holding a permanent residence there.
The survey uses a three‑stage random sampling design, consisting of municipalities (stage 1), institutions in the homelessness-assistance sector and adjacent support systems (stage 2), and people experiencing homelessness (stage 3). First, a random sample of municipalities is drawn. Second, within the selected municipalities, institutions that may be in contact with people experiencing homelessness are identified to construct the sampling frame using administrative data, from which a random sample of institutions is drawn. The service network covers a broad range of institutions, including day centres, counselling centres, street outreach services, job centres and youth services.3 Third, within the participating institutions, people experiencing homelessness are randomly selected for the survey (OECD, 2026[4]).
The survey operates under the assumption that nearly all people experiencing homelessness have some contact with the support system. The sample is constructed to allow extrapolation to the national population of both ETHOS Light categories (e.g. people sleeping rough and people staying with family or friends). In 2024 and 2026, the sampling of municipalities was expanded from previous years to improve data quality and extrapolation accuracy – and, in 2026, to improve regional coverage.4
The survey is conducted over a fixed one‑week window (1‑7 February) during which the selected institutions complete counting protocols and surveys that include more detailed questions to people experiencing homelessness. To verify the results against the annual reporting of sheltered homelessness and avoid potential double‑counting across both data collection approaches, respondents are asked where they stayed on the night of the 31 January – the day on which the sheltered survey is conducted. In 2024, 906 municipalities across 200 municipalities surveying 1 556 unsheltered homeless people and 694 staying with family or friends. This represented a response rate of 49% of the sampled institutions. DIFU works with a subcontractor to analyse the questionnaires, as this involves several time‑consuming steps, including sorting, recording, stamping and digitising5 (Bundesministerium für Wohnen, Stadtentwicklung und Bauwesen, 2024[9]).
The methodological design of Germany’s approach has partially evolved over time. The first biennial report included a feasibility study on the use of additional data sources, such as register and census data, to complement the service-based approach (Bundesministerium für Arbeit und Soziales, 2022[8]). Building on this work, the 2026 edition piloted the use of administrative data from job centres and social welfare offices. Preliminary findings suggest that this approach identified more people staying with friends or family, suggesting that administrative register could help to improve the measurement of ETHOS Light 6. It may also provide prevalence estimates that are close to the actual scale of homelessness while reducing the data-collection effort.
However, the use of administrative data also comes with important limitations. It relies on proxy indicators of homelessness, such as the absence of housing cost payments and address registration with an independent welfare organisation. Even where such criteria are met, the presence of homelessness cannot be established with certainty. In addition, differences in how institutions interpret data protection requirements affected their willingness to share data. These findings underline the need for clear and consistent frameworks for data protection, data sharing and data processing if administrative data is to play a larger role in homelessness measurement.
Strengths and limitations
Germany’s approach to measuring homelessness combines methodological breadth with a legally anchored framework to facilitate continuous methodological improvement. The biennial survey relies on an extended service-based approach that covers a broad and heterogeneous network of institutions, which increases the likelihood of capturing both the unsheltered homeless population and people staying with family or friends, including individuals who are typically missed or undercounted because they do not access homelessness services (see Section 2.2.1). Germany’s approach aligns with the extended service-based method established in the Nordic countries (Benjaminsen et al., 2020[10]), as illustrated in the next country example (Denmark). The random sampling design with national extrapolation is adapted to Germany’s size, where a full national enumeration would prove difficult. Notably, thanks to the explicit question in both surveys about the living situation of the respondent on 31 January, the annual sheltered-count and biennial survey are explicitly designed to produce a combined national homelessness estimate while estimating potential double‑counting.
Further, the legal framework for data collection builds in a staged, institutionalised framework for methodological development, including a feasibility study in the first version and an expansion assessment in the third. This signals a structural commitment to improving coverage of homelessness over time, as demonstrated in practice by the 2026 pilot of complementing the service-based approach with administrative data.
However, there are several limitations. Both reports acknowledge that an unknown degree of undercounting persists. First, the service-based, point-in-time approach can only capture people who have contact with the support system during the survey week. This is a typical shortcoming of survey-based approaches. People staying with family or friends are especially at risk of being undercounted, as they are more likely to have a limited and less predictable relationship with formal services, and are not necessarily in contact with services during the survey week. In addition, both reports note the significant underrepresentation of people who may be experiencing homelessness in eastern German states relative to population size. This is attributed at least in part to lower service density, rather than lower prevalence of homelessness. To address this challenge and improve regional coverage, the 2026 survey adjusted the methodology with a higher municipal sample in eastern Germany.
Further, people experiencing homelessness are surveyed directly by service providers, and participation is voluntary. Given the sensitive nature of questions relating to housing history, health and immigration status and experiences of violence, selection bias cannot be ruled out.
Finally, there are several important methodological limitations. First, the extrapolation approach involves calibration decisions that can produce substantial variation in headline figures. Changes in weighting assumptions can significantly affect the estimates. Second, the relative complexity of the survey instrument, which involves counting protocols, detailed questionnaires and a separate digitisation and analysis process that requires a subcontractor, creates operational burdens that may affect institutional response rates and introduce continuity risks. Changes in the commissioned research institutes from one survey edition to the next also bear the risk of efficiency losses.
Denmark: A service-based biennial point-in-time survey
Denmark’s national estimate of homelessness is based on a service-based biennial point-in-time survey, Mapping of Homelessness (OECD, 2026[4]).
Institutional context and objectives
Mapping of Homelessness is a point-in-time survey conducted once every two years by an independent research institute (VIVE, the Danish Centre for Social Research), in collaboration with Statistics Denmark. The study is funded through the Development and Investment Programme under the Ministry of Social Affairs (VIVE, 2024[11]),6 with an allocation of DKK 1.2 million (around EUR 160 500) to VIVE for a two‑year period in 2025. From 2027 onwards, the government will allocate DKK 0.7 million (around EUR 93 500) annually to VIVE to fund the mapping including the biennial homelessness count.7
The Mapping of Homelessness survey exists since 2007.8 The survey was initiated in parallel to the development of a national strategy to address homelessness in the context of welfare reforms that coincided with an increase in youth homelessness and underlined a growing concern for marginalised groups (Dyb et al., 2021[12]). The purpose of the survey is to document the extent, nature and development of homelessness nationally and within individual municipalities, including the socio-demographic profile of people experiencing homelessness to inform action in the policy area (VIVE, 2024[11]). The mapping is considered a vital tool to enable policymakers to steer national homelessness policy (Benjaminsen et al., 2020[10]). There is no legal framework that governs the data collection of homelessness in Denmark (OECD, 2026[4]).
Data collection approach and methodology
Denmark’s homelessness survey relies on a service-based method and is carried out in calendar week 6 of the respective survey year. The homelessness mapping distinguishes eight homelessness situations, which generally correspond to ETHOS Light categories 1, 2, 3, 4 and 69 (OECD, 2024[13]). This reflects a broad conception of homelessness and housing exclusion, which is typical to Nordic countries (Benjaminsen et al., 2020[10]). Several clarifications on specific ETHOS Light categories in the Danish context are worth noting:
The Danish operational definition corresponding to ETHOS Light 1 is defined as “nights on the street, in the stairwell, in a tent” (VIVE, 2024[11]).
ETHOS Light 5 is not recorded as a separate category. Instead, around 6% of individuals are registered under “other” forms of homelessness, such as stays in garden houses or caravans, which the ETHOS typology would usually classify as category 5.
ETHOS Light 6 – referred to in the Danish report as sofa surfing – includes temporary stays with family or friends, as well as situations in which individuals stay overnight with “acquaintances in an abusive environment characterised by high instability” (VIVE, 2024[11]).
The mapping is implemented in two stages: first, the identification of relevant services and authorities, and second, the completion of individual questionnaires by these actors during the census week.
Identification of relevant services and authorities
The survey relies on a broad network of services, including those outside the homelessness sector, and can thus be classified as an “extended service-based count” (Benjaminsen et al., 2020[10]). All social services, municipal authorities and institutions that may have contact with or knowledge of people experiencing homelessness are invited to contribute. This includes many actors for whom homelessness is not the primary target group, such as job centres, citizen service and benefit units, the police and probation services, psychiatric services, addiction treatment facilities, and warm rooms or counselling services provided by NGOs.
The list of participating units is updated by VIVE through consultation with the previously involved social services and authorities. All 98 Danish municipalities are included in the count and no sampling or weighting procedure takes place. In 2024, 1 421 units of services and/or authorities received information about the mapping with a request to distribute the information to all employees who may have contact with or knowledge of citizens experiencing homelessness.
Completion of individual questionnaires
In the second step, staff members of the relevant service providers and authorities complete a two‑page form for each individual with whom they have been in contact or know to be in a situation of homelessness during week 6. In addition to the specific homelessness situation, the form also requests information relating to demographic characteristics, physical and mental health conditions, substance use, income source, duration of homelessness, pathways into homelessness and ongoing support interventions. The form can be completed with or without the involvement of the individual concerned. In practice, joint completion occurs in only around 5% of cases. The data collection is permitted through national data protection laws and by special permissions in the case of health authorities for data collection of national interest (Benjaminsen et al., 2020[10]). The form may be completed electronically or on paper. The electronic completion is encouraged and takes place via a digital reporting platform provided through Statistics Denmark. Statistics Denmark is responsible for distributing the material and collecting the data, whereas VIVE is responsible for data processing and analysis. Submission of the form is possible until 1 March each year.
Because individuals may be in contact with more than one service, a unique identifier is used to prevent double counting. In Denmark and Sweden, data law permits collecting the full personal identification number (Benjaminsen et al., 2020[10]). Where this unique identifier is unavailable, date of birth or initials may serve as substitutes. If several forms exist for the same individual, priority is given, where relevant, to forms completed with the participation of the person experiencing homelessness and to the most complete questionnaire. In the 2024 survey, 1 500 cases of double counting were identified (VIVE, 2024[11]).
The response rate of the 1 421 units participating in the survey in 2024 was approximately 73%, down from 85% in 2017, continuing a downward trend in recent survey editions. Institutions that have not submitted a response are reminded via e‑mail or phone. Strategically, reminders are prioritised with personal contacts to the most important respondents in larger municipalities and who report a major share of the overall homelessness estimate (Benjaminsen et al., 2020[10]). Despite the slight decrease in response rates, the data are considered comparable over time, as non-responding units often have little or no contact with people experiencing homelessness (this is the case for around half of participating units). Overall, the degree and type of uncertainty is therefore assessed as relatively stable (VIVE, 2024[11]).
It is worth noting that compared with the 2022 survey, people staying with family or friends increased by more than 25%, whereas the total numbers across categories rose only marginally. Part of the increase may reflect improved municipal awareness following legislative changes introduced shortly before the 2024 mapping (VIVE, 2024[11]). These reforms altered reimbursement rules that may have strengthened incentives for municipalities to secure more stable housing solutions for people experiencing homelessness and overall attention to the issue.
Strengths and limitations
The Danish mapping illustrates several strengths of an extended service-based approach. First, including a broad service network increases the probability of capturing heterogeneity of homelessness experiences. Second, the knowledge‑based reporting approach – whereby staff can provide information about individuals they know of, even without direct contact during the census week – can also help improve the measurement of ETHOS Light 6, since this group is not necessarily in contact with services every week, even where the service network is very broad.
Allowing staff to complete forms on behalf of people experiencing homelessness could reduce the selection bias that might otherwise arise if participation depended on an individual’s participation in completing the form (which can be more difficult, particularly for individuals experiencing unsheltered homelessness) – a concern that has been raised by homelessness advocacy organisations in Germany. However, given that most forms in Denmark are completed by staff members rather than by people experiencing homelessness themselves, some responses reflect staff assessments or are left unanswered in cases where staff are simply unaware of the relevant information (VIVE, 2024[11]).
Additionally, the data generated by the count is used both to design homelessness programmes and to monitor their performance in the long-term, including the identification of priority groups through the survey’s detailed profiling (Benjaminsen et al., 2020[10]). Because more than four in five registrations contain the full personal number, these records can be linked to general register data under strict safety procedures, enabling more detailed analysis of risk factors and pathways into and out of homelessness (Benjaminsen et al., 2020[10]). Such a concrete policy link can help boost motivation for participation in the mapping. Municipal awareness and engagement with the mapping also appear to be reinforced by the funding and reimbursement rules tied to the policy areas.
There are nevertheless limitations. First, the point-in-time design cannot capture how many people experience homelessness over a longer period. VIVE estimates that 12 000‑15 000 individuals may be affected over the course of a year, which is more than double the point-in-time headline estimate (VIVE, 2024[11]). Second, as with any service-based method, individuals are only enumerated if they are known to a participating institution (OECD, 2025[2]). The resulting figure should therefore be read as a minimum estimate. Undercounting is particularly likely where response rates among service providers are lower – for instance in smaller municipalities, or among providers that do not see themselves as part of the social sector, such as job centres. VIVE acknowledges that this may limit the extent to which more hidden forms of homelessness can be captured through the survey (VIVE, 2024[11]).
Further, response rates by services and authorities have declined in recent editions, possibly reflecting growing competition from other surveys as well as administrative reporting burdens on local authorities – a challenge also faced by Sweden and Norway, which use a similar approach (Benjaminsen et al., 2020[10]). Even relatively small changes in participation observed in recent years are reflected in lower headline numbers that do not necessarily reflect a real change (Benjaminsen et al., 2020[10]). While declining response rates could in principle be addressed through sampling and weighting, this also carries a risk of distortion (Benjaminsen et al., 2020[10]).
Lessons from service-based methods
service-based methods are the most widely used approach for measuring both ETHOS Light 1 and ETHOS Light 6. In Denmark and Germany, the network used to measure homelessness is extensive, deliberately including institutions whose primary target audience is not explicitly for people experiencing homelessness, such as employment agencies. This improves the coverage of heterogeneous groups of people experiencing homelessness.
Nevertheless, this approach cannot overcome the inherent methodological limitation of service-based methods, which is that they systematically exclude people who are not in contact with services. The country case studies demonstrate that several key design choices can be approached differently when implementing the method and are likely to affect the results. For instance:
Full enumeration vs. sampling and weighting: Denmark conducts a full enumeration by sending the survey to all municipalities, whereas Germany extrapolates the overall national estimate from a three‑stage sampling process. While sampling and weighting risks distortion in the data, it presents a good option where a full enumeration appears unfeasible.
Strict point-in-time vs. knowledge‑based reporting: Germany’s approach involves a stricter implementation of a point-in-time count, whereby only people experiencing homelessness are counted in the survey week. In contrast, staff at Danish institutions in contact with people experiencing homelessness are asked to report anyone they know, which is likely to improve coverage of population groups not in regular contact with services. In the European Homeless Count pilot project (see Box 2.1), staff at service providers can complete a questionnaire for individuals experiencing homelessness with whom they have not been in contact on the day of the count. However, this contact must have taken place relatively recently (i.e. within two weeks), and the person’s sleeping location must be verified (Teller et al., 2025[1]).
Staff-reported vs. self-reported information: In Denmark, most individual forms are completed by staff in contact with the person experiencing homelessness, without necessarily involving the individual, whereas in Germany, people experiencing homelessness are surveyed directly. The Danish approach may reduce self-selection bias in the survey results, but it can also lead to incomplete information.
Counting through service providers only vs. complementing with administrative data: In the third edition of its biennial survey, Germany trialled the use of administrative data from job centres and social welfare offices, thus complementing a service-based survey approach with administrative data. While this seems to have the potential to provide a more accurate estimate of the actual prevalence with comparatively lower data collection efforts, it also presents new implementation challenges such as defining proxies and assumptions and ensuring data sharing across institutions.
Administrative data as a basis for measuring unsheltered homelessness and temporary stays with family or friends
Copy link to Administrative data as a basis for measuring unsheltered homelessness and temporary stays with family or friendsAdministrative data refers to records and registries maintained by various government institutions and service providers, such as social benefits systems, health services, criminal justice records, and emergency shelter databases, which are repurposed to estimate the number of people experiencing homelessness, even though the data were not originally collected for this purpose. The approach can in principle be tailored to cover all six ETHOS Light categories, depending on which registries are drawn upon and how they are linked; in addition, administrative data can generate either point-in-time snapshots or flow data over a period (OECD, 2025[2]).
The main strengths as outlined in the OECD Monitoring Framework to Measure Homelessness (2025[2]) are that the data already exist and can be frequently collected. Further, administrative data can provide detailed demographic and personal information when registries are cross-referenced, which can support individualised monitoring and case‑level follow-up, as well as more tailored prevention measures.
However, among the main limitations are that administrative data only capture people who are in contact with formal services or the welfare system; this means that anyone who avoids or lacks access to these systems is not covered. This represents a particular challenge for homelessness, given that many people experiencing homelessness have limited, irregular, or no access with formal service systems. Similar to service-based methods, this can lead to underestimating specific population groups that are not (accurately) represented in administrative records, such as undocumented migrants. Generally, coverage can be weaker in areas with low service availability or take‑up. Data quality and completeness can vary across registries; and combining datasets across institutions raises practical challenges around privacy, data protection legislation, and the resources and technical expertise needed to harmonise different systems.
Finland: Annual municipal reporting from administrative data sources
The Centre for State‑Subsidised Housing Construction (Varke) in Finland collects national data on homelessness annually through surveys circulated to municipalities, which consolidate and report data using a variety of administrative data sources.
Institutional context and objectives
Finland’s national estimates of homelessness are based on data collected annually from municipalities by Varke, an independent state agency established in March 2025 as part of the Ministry of the Environment, succeeding the Housing Finance and Development Centre of Finland (ARA). There is no legal framework governing the collection of homelessness data in Finland: data collection is voluntary (Centre for State-Subsidised Housing Construction, 2026[14]). A precise estimate of the total cost of the annual mapping cannot be provided, since Varke does not have a specific budget line for it and data collected relies on existing administrative sources. Most of the work involved in the mapping takes place in municipalities and in social services, with considerable variation depending on the size of the municipality, the number of people experiencing homelessness and the number of data sources used.
Finland was the first Nordic country to produce national homelessness estimates, with data collected since 1986, though following a fundamentally different methodology than the other three Nordic countries (Benjaminsen et al., 2020[10]). As in other Nordic countries, the regular surveys form a core part of knowledge production within Finnish homelessness policy and function as a central steering mechanism, providing the central government with a tool to monitor developments and assess the effectiveness of policies to prevent and reduce homelessness (Benjaminsen et al., 2020[10]). Notably, the national government’s formal steering tools are limited, and municipalities implement homelessness policy with a high degree of autonomy. Consequently, it is important for municipalities themselves to see a benefit in the statistics for their local work (Dyb et al., 2021[12]).
Data collection approach and methodology
The operationalisation of different types of living situations that correspond to homelessness in Finland is explicitly aligned with the ETHOS Light typology and distinguishes four categories (OECD, 2026[4]):
People living outdoors, in stairwells, or in overnight shelters that offer only a single night (Finland includes this category under ETHOS Light 2 with no separate breakdown for ETHOS Light 1).
Dormitories or accommodation businesses (feeding into ETHOS Light 3).
Housing service units, rehabilitation units, hospitals or other institutions not primarily intended for people experiencing homelessness, including shelters for victims of domestic violence (corresponding to ETHOS Light 4).
People staying temporarily with friends or relatives due to lack of housing (corresponding to ETHOS Light 6). Young people under 25 living with their parents are explicitly excluded from this category.
The survey is a one‑day point-in-time count, last conducted on 15 November 2025. Varke circulates an electronic survey to all 292 municipalities in mainland Finland in October with instructions on how to complete it. In 2025, 238 municipalities responded (81%): 99 municipalities reported people experiencing homelessness; 139 municipalities reported none; and 54 (predominantly small) municipalities did not respond to the survey. The non-responding municipalities are individually listed in the report’s appendix. Rather than collecting individual questionnaires directly from service providers as in service-based methods, municipalities directly compile data from one or more of the following local administrative sources:
social services registers
housing applicant registers of municipal rental housing companies
customer registers of service providers
Social services registers are the most commonly used administrative data source. However, the establishment of well-being services counties in 2023, which are now the legal holders of social services registers, has introduced operational complications: municipalities remain responsible for reporting to Varke, but must now request access to data held by their respective well-being services county, raising questions around data processing responsibilities and arrangements.
The registries that record housing applicants of municipal rental housing companies are an important additional source of data, particularly in larger municipalities, where some have found that up to 40% of people experiencing homelessness are not in contact with social welfare services at all.10 Municipalities are encouraged to use the Digital and Population Data Services Agency’s population register (VTJ) to verify the housing status of housing applicants. Because homeless status is prioritised in social housing allocation, some individuals may misreport their housing situation, and checking against the population register can reveal that a person in fact has a permanent address (Benjaminsen et al., 2020[10]). However, using the population register requires case‑by-case consideration, as it is known that the register information of people experiencing homelessness is not always up to date.
Since the centralisation of social assistance to Kela in 2017, the ten largest municipalities receive a list of social assistance applicants identified as homeless directly from Kela for the month of October. Municipalities are encouraged to use the population register data to verify these estimates.
Municipalities cross-check data from different sources using social security numbers to remove duplicates, and report to Varke only the total figures of people experiencing homelessness in the different groups. No personal information is transferred to Varke, and thus Varke has no means of verifying who was counted as homeless. Notable anomalies, such as a stark rise or fall in the number of homeless people compared to previous year, can be re‑examined with the relevant municipality (Centre for State-Subsidised Housing Construction, 2026[14]). The processing of personal data for homelessness enumeration is grounded in the Data Protection Act (§4), which permits processing where necessary for a task carried out in the public interest, supplemented by the Openness Act (§28) as a basis for data disclosure (Centre for State-Subsidised Housing Construction, 2025[15]). Only aggregate figures are forwarded to Varke (Centre for State-Subsidised Housing Construction, 2025[15]).
Strengths and limitations
The extended time series allows the report to track trends and major changes in homelessness over the long term, providing a strong basis for monitoring municipalities’ homelessness work and for planning services and housing provision. The report includes municipalities’ own assessments of how their figures have developed (Centre for State-Subsidised Housing Construction, 2026[14]). The response rate is high at 81%, which experts at Varke partly attribute to institutional familiarity given that the survey has been conducted since 1986, as well as to considerable public and political attention on the topic. The fact that the receipt of certain municipal grants is conditional on participation may also support reporting rates.
Regarding estimates for ETHOS Light 6 specifically, which account for the majority of the overall homelessness figure in Finland (64% in 2025), the Finnish case demonstrates the potential of mobilising administrative data sources beyond those from social services alone. As noted above, some municipalities report that up to 40% of people experiencing homelessness were not in contact with services at all, making the use of housing registers critical for identifying this population.
Compared to the extended service-based counts in Denmark, Norway and Sweden, no individual questionnaires are completed. Instead, information is gathered at the level of population groups using a more restricted set of demographic characteristics, resulting in considerably less profiling information overall (Benjaminsen et al., 2020[10]).
Methodologically, the Finnish data carries the typical limitations of a point-in-time count, capturing only a snapshot rather than homelessness over a longer period. The reliance on municipal self-reporting also introduces comparability challenges: given that data collection has no statutory basis, Varke cannot enforce a uniform approach to it. Consequently, municipalities’ information acquisition methods and assessment criteria vary, and municipal data collection approaches are not always well documented (Centre for State-Subsidised Housing Construction, 2026[14]). Municipalities’ assessments of the reliability of their own figures also vary considerably, with some reporting high confidence in precision and others considerable uncertainty. Consequently, there is no quantitative estimate of the magnitude of undercounting at the national level. Experts from Varke assume that young people and people with migration background are the population groups most likely to be undercounted.
Given that municipalities are encouraged to draw on multiple registers, for which access (at least with respect to social service registers) requires active co‑ordination with additional administrative level (e.g. well-being services counties), the Finnish data collection process is less straightforward than a point-in-time administrative data pull might suggest. More detailed guidance on data processing and protection in light of the changes to register-holdership structure would be beneficial. This organisational complexity, combined with some municipalities’ view that reporting responsibility should shift to the well-being services counties, could also contribute to non-response.
Considering further development of Finland’s measurement approach
In terms of efforts to develop the approach, Statistics Finland investigated in 2023 whether a homelessness survey could be conducted on the basis of national register data alone, using a process of elimination applied to population registers to identify individuals without a permanent registered address. The study concluded that reliable figures could not be produced, primarily because the two most critical data sources – social welfare client registers and municipal housing applicant registers – were inaccessible due to data sharing constraints (Statistics Finland, 2023[16]). A more recent development is that the Finnish Institute for Health and Welfare is currently working to consolidate well-being services county data into a single national register, which could in the future be leveraged for homelessness measurement. There is, however, agreement that social service register data alone would need to be complemented by other sources, given the high share of people experiencing homelessness who are not in contact with services.
Lessons from administrative data
Administrative data can improve coverage of less visible forms of homelessness, specifically ETHOS Light 6. This is evidenced by the high rates of ETHOS Light 6 reported in Finland, as well as by the preliminary insights from the German pilot (Section 3.2.1). This is because it can be assumed that people experiencing homelessness are more likely to be captured in some registers or administrative databases than accessing services on a weekly basis – this, however, being the underlying assumption of a strict point-in-time, service-based counting approach.
Still, reliance on a single register or administrative database is unlikely to provide a reliable estimate of homelessness. Instead, combination or verification through multiple sources is required, as demonstrated by Finland’s decades-long experience. This implies that the process of relying on administrative data for homelessness measurement can be less straightforward than a point-in-time administrative data pull might suggest. Moreover, homelessness figures derived from administrative data sources that were not initially designed for measuring it depend on the definition of proxy indicators and criteria, which will inherently affect the results and their reliability. Lastly, accessing and processing data, and the governance frameworks linked to this process, emerge as key implementation challenges, as demonstrated by changes to social assistance register ownership in Finland.
Population census and household surveys as a basis for measuring unsheltered homelessness and temporary stays with family or friends
Copy link to Population census and household surveys as a basis for measuring unsheltered homelessness and temporary stays with family or friendsPopulation censuses and household surveys can allow for the collection of in-depth information about different forms of homelessness, particularly including people staying with family or friends (OECD, 2025[2]). Depending on the design of questionnaire, the collected data may include information about individuals’ demographic characteristics, their experience(s) of homelessness, and/or their support needs. Typically, population censuses and household surveys cover ETHOS Light categories 2, 3 and 6; information about individuals’ past experiences of homelessness can also be captured. When combined with a street count conducted on the same day, a population census can additionally cover ETHOS Light categories 1 and 5, offering a relatively comprehensive overview of various experiences of homelessness.
Population census and household surveys do not, however, provide directly actionable data for policymakers, as the exercises are undertaken either too infrequently (census), or with inadequately sized samples (household surveys). Moreover, the time lag between the data collection and the release of the results undermines the timeliness of the findings. Finally, since censuses and surveys are typically completed by the head of the household, certain experiences of homelessness, such as youth homelessness, may not be well captured.
Australia: Point-in-time count of the Census of Population and Housing
Australia’s national estimates on homelessness are based on the Census of Population and Housing, a point-in-time count conducted in August every 5 years by the Australian Bureau of Statistics (ABS).
Institutional context and objectives
Australia’s most recent census was conducted in August 2026, with the most recent available Census data based on the 2021 Census. Total costs for the 2021 census were approximately USD 565 million, or about USD 22 per person (Australian Bureau of Statistics, 2022[17]) – notably lower than, for example, the New Zealand Population Census, which also enumerates the population experiencing homelessness (Stats NZ, 2025[18]). There is no specific legal obligation relating to data collection for people experiencing homelessness, but all individuals in Australia are obliged to respond to the Census, including people experiencing homelessness (OECD, 2026[4]).
Since 2011, the census has incorporated a dedicated homelessness enumeration strategy. Data on the number and location of people experiencing homelessness supports the targeting, planning and funding of services to meet their needs (Australian Bureau of Statistics, 2021[19]). In 2012, the ABS Homelessness Statistics Reference Group published the method to derive homelessness estimates from the Census of Population and Housing, and this framework has been applied consistently since, with minor changes for improved measurement (Australian Bureau of Statistics, 2012[20]). Consequently, data are comparable across Census years (OECD, 2026[4]).
Data collection approach and methodology
The Census’ homelessness enumeration strategy targets three streams of people experiencing homelessness on the night of the Census:
people not in a dwelling (e.g. improvised dwellings, tents or sleeping out)
people in a private dwelling (e.g. staying temporarily with other households or living in severely crowded dwellings)
people in non-private dwellings (e.g. supported accommodation, boarding houses or other temporary lodging)
Together, these living situations roughly correspond to ETHOS Light categories 1, 2, 3, 5 and 6 (OECD, 2026[4]). Some subcategories typically associated with ETHOS Light 5 (such as individuals staying in improvised dwellings and tents) are included under ETHOS Light 1 in the Australian context.11 In addition, people staying with family or friends (ETHOS Light 6) are included in a broader category of people experiencing homelessness in a private dwelling, which also includes people living in severely overcrowded dwellings. The statistical definition of homelessness in Australia is thus among the most encompassing in OECD/EU countries.
Census materials are distributed either by mail or through household visits, and forms can be completed online or returned in paper format. In addition, administrative information from governments and homelessness services is used to help identify the location and characteristics of the homeless population. To strengthen coverage, the ABS implements a Homelessness Engagement Strategy aimed at ensuring effective collaboration across jurisdictions and service providers and at recruiting field staff with on-the‑ground experience with people experiencing homelessness and with specific communities (Australian Bureau of Statistics, 2021[21]).
Measuring ETHOS Light 1 and 5
For people not in a dwelling (ETHOS Light 1 and 5), over several days commencing from the night of the Census, specialised staff interview individuals in locations that deliver services for people experiencing homelessness.12 Community Field Officers require a certain level of expertise to ensure an accurate count, with targeted recruitment from the homelessness sector or with lived experience. People sleeping rough are interviewed using a shortened version of the standard form (Australian Bureau of Statistics, 2021[21]). After identifying all individuals enumerated in improvised dwellings, tents or sleeping out, additional indicators, such as tenure, income, rent or mortgage payments, and labour-force status are applied to exclude persons who are likely to have alternative accommodation and therefore would not be considered homeless (Australian Bureau of Statistics, 2012[20]).
Measuring ETHOS Light 6
Most people in private dwellings complete the census online or on paper. This includes individuals staying temporarily with other households (ETHOS Light 6). The ABS first identifies everyone in private dwellings who reported having no usual address, which is defined as the address at which a person has lived or intends to live for six months or more. As there are many reasons why a person may not currently have a residence without being de facto homeless, such as extended travel or recent mobility, a number of groups with likely accommodation alternatives are subsequently excluded. To this end, the ABS applies a multi-step exclusion logic, drawing on dwelling type, tenure and ownership, labour force status, age, and country of birth and migration history to systematically remove persons likely to have accommodation alternatives, such as retirees who are travelling for extended periods (“grey nomads”) or residents who only returned to Australia in the year of the Census after a period of emigration.
The remaining population is considered, on balance, to lack tenure and control over space and therefore to meet the homelessness definition (Australian Bureau of Statistics, 2012[20]). For the 2021 census, the wording of the usual-address question was revised to include examples such as family conflict or eviction as valid reasons for reporting “none” (Australian Bureau of Statistics, 2021[21]).
Strengths and limitations
From a methodological perspective, Australia’s census-based approach has several notable strengths. First, it provides relatively comprehensive and inclusive coverage of homelessness, capturing people across a wide range of living situations. The broad statistical definition is one of the most encompassing in OECD countries. Second, there is relatively strong consistency and comparability over time, with the same method applied since 2012 (with only minor refinements). Third, Census data are supplemented with administrative information and a dedicated engagement and enumeration strategy that leverages collaboration with homelessness services, local stakeholders, and experienced field staff, while continuously refining targeted approaches for population groups at greatest risk of undercounting across successive Census cycles.
There are nonetheless several limitations. As with census-based methods in general, this approach does not provide policymakers with timely insights on homelessness due to the infrequent data collection and long publication delay (OECD, 2025[2]). Because census variables were designed for general population purposes, they must be interpreted as proxies when used to estimate homelessness. This comes with the risk of under- or over-estimating homelessness figures. People experiencing unsheltered homelessness are particularly vulnerable to being underestimated, which is why the ABS has developed a dedicated enumeration strategy in co‑operation with service providers. Although a Census Post-Enumeration Survey exists to assess general coverage, it cannot be used to estimate undercounting of people experiencing homelessness because it does not gather sufficient information about homelessness status at the relevant time.
The ABS highlights that certain population groups, often those who are temporarily staying with family or friends (ETHOS Light 6), are more strongly affected by undercounting (Australian Bureau of Statistics, 2012[20]), notably:
People displaced by domestic or family violence are likely to be undercounted. Some may be reluctant to report having no usual address due to stigma or because they assume they can return to their former residence.
Young people in ETHOS Light 6 are particularly difficult to identify. They may retain a registered address, may not wish to disclose their actual situation where they are staying, or another household member may complete the form assuming that they will return home.
Aboriginal and Torres Strait Islander peoples are more likely to be under-enumerated and over-represented in homelessness statistics. Even where counted, homelessness may be underestimated because cultural understandings of home and mobility do not always align with the concept of having “no usual address”.
Czechia: Estimates for people staying with family or friends, which are extrapolated from household survey data
In Czechia, official national estimates on homelessness and housing exclusion are derived from the Census of People from Selected Categories of the ETHOS Classification, most recently conducted in 2025.
Institutional context and objectives
The ETHOS Census was first carried out in 2019. Although there is no legally established fixed cycle requiring data collection on homelessness, the Census has so far been conducted every three years, and efforts are underway to secure this frequency on a permanent basis (OECD, 2026[4]). The defined objective of the Census is to obtain qualified estimates of the number of people experiencing homelessness and housing exclusion across the entire national territory (Král et al., 2025[22]). The Census pays particular attention to improving the reporting of individuals who are difficult to detect and to capturing persons living in insecure and inadequate housing (OECD, 2026[4]). The project is commissioned by the Czech Ministry of Labour and Social Affairs and implemented by SocioFactor, a private research company specialising in applied social sciences (OECD, 2026[4]). The total contract value amounts to CZK 4 900 000 excluding VAT (approximately EUR 200 000).13
While described as a census-based approach, the point‑in‑time enumeration draws on a range of complementary methodological approaches, including street counts, service-based methods, administrative data, and household surveys, which together allow for the inclusion of all ETHOS Light categories (OECD, 2026[4]). The Czech methodology adapts the ETHOS framework to national conditions, while retaining its four conceptual categories (Král et al., 2025[22]):
people without a roof,
people without a dwelling,
people in inadequate housing,
people in insecure housing.
Data collection approach and methodology
In the Czechian context, individuals temporarily staying with family or friends (ETHOS Light 6) are considered under the broader operational category of people experiencing insecure housing. Such arrangements are typically temporary, lack a contractual relationship, often involve restrictions on privacy (e.g. staying in a partner’s home after separation), and may precede other forms of homelessness (Král et al., 2025[22]). Estimates are produced by extrapolating European Union Statistics on Income and Living Conditions (EU-SILC) micro‑samples to the Czech population. Experts were consulted to interpret and scale the EU-SILC findings in order to derive estimates for the prevalence of ETHOS Light 6 (Král et al., 2025[22]).
To estimate ETHOS Light 6, it is essential to distinguish between voluntary cohabitation – which is far more common – and forced cohabitation. Since no specific EU-SILC question directly captures whether cohabitation is voluntary or necessity‑driven, individuals are counted in this category only when all of the following conditions, which were defined in consultation with experts, are met (Král et al., 2025[22]):
The household is staying with relatives or friends.
Financial distress, identified where a household meets at least one of the following criteria:
has great difficulty making ends meet,
cannot afford an unexpected expense, and
falls below the income poverty line (60% of the median income).
The household occupies a dwelling with 8‑16 square metres of floor space per person.
Strengths and limitations
Reliance on a household survey such as EU‑SILC can promote the inclusion of people experiencing homelessness who may remain hidden in other methods to measure homelessness, specifically if they do not access shelters or services. This is particularly relevant for households that avoid contact with service providers or do not self‑identify as experiencing housing exclusion and homelessness. Further, the Czech approach mobilises a survey that already exists, which might bring efficiency gains.
However, the ETHOS Census report identifies several limitations of the data collection approach (Král et al., 2025[22]). Generally, there is a risk of underestimation, as households may choose not to disclose their circumstances in survey settings. Among the large number of people living in a shared household, estimated at approximately 650 000 in Czechia, the key methodological challenge – as outlined above – is to distinguish necessity‑driven cohabitation from voluntary arrangements, and agreeing on the criteria upon which to base this distinction. For example, it is debatable whether overcrowding must always be present to classify as a living situation corresponding to ETHOS Light 6. This challenge illustrates the limitations of using survey data not specifically designed to measure homelessness, as estimates necessarily depend on a series of assumptions, which can significantly influence the results. This is demonstrated by the significant decrease of ETHOS Light 6 between 2022 and 2025, which likely stems from methodological changes (the 2025 estimate applied narrower criteria to define necessity‑based co‑residence), rather than a real decline in homelessness among this group.
Overall, there is substantial uncertainty in the estimate. The confidence interval for this group is ±50%, meaning that the true figure could be significantly lower or higher. This uncertainty arises because the EU-SILC sample size is too small to produce a robust estimate for this specific subgroup. Inadequate sample size is a key limitation of household surveys as a method to measure homelessness (OECD, 2025[2]).
Lessons from population censuses and household surveys
Population censuses and household surveys can provide estimates of a broad range of living situations – including more hidden forms of homelessness, such as people staying with family or friends – without relying on individuals to be in touch with service providers. However, as they are based on self-reporting by individuals, they are likely to undercount or exclude certain groups, such as young people, whose survey forms may be filled out by another household member, or victims of domestic violence, who may provide the address of their usual residence due to fear of stigma or the assumption that they can return home.
As population and household surveys are not necessarily designed to measure homelessness, they rely heavily on proxy variables based on different assumptions to determine whether individuals are staying with family or friends due to a lack of alternatives. This is evident across the relevant country case studies Australia and Czechia. Such assumptions affect results and, by extension, the potential reliability of the results. Household surveys are further constrained by their relatively small sample size, which can lead to high uncertainty in the results. This is demonstrated by the ±50% confidence interval in the Czechia’s ETHOS Light 6 data, which was extrapolated from EU-SILC. Lastly, measurement approaches that rely on population censuses produce infrequent estimates, which does not allow for timely policy insights.
Information management systems as a basis for measuring unsheltered homelessness and temporary stays with family or friends
Copy link to Information management systems as a basis for measuring unsheltered homelessness and temporary stays with family or friendsInformation management systems, which integrate data from multiple service providers or local authorities into a single, coherent system for a given geographic area, constitute an important method for collecting data on homelessness. Once established, these systems can provide a cost-effective means of generating detailed and dynamic data, as they reduce the duplication of data collection. They typically operate at client level, enabling real-time, personalised information that supports targeted and co‑ordinated service delivery and monitoring of cases. The data can also facilitate programme evaluation (OECD, 2025[2]).
However, establishing such systems is usually resource‑intensive and requires technical expertise, as well as consistent data practices across providers. Such requirements can be difficult to meet in a context of budgetary or staffing constraints. Because the information recorded is client‑level, privacy and data‑protection are also key concerns. These can sometimes be mitigated using unique identifiers, though in smaller communities even pseudonymised data may not be sufficient to ensure full anonymity (OECD, 2025[2]).
United Kingdom (Scotland): Information management system
The official statistical estimate on homelessness in the United Kingdom (Scotland) is provided by the Homelessness in Scotland bulletin (OECD, 2024[13]). This flow count is based on information gathered from individuals presenting to local authorities for homelessness support, and the data collection can therefore be classified as an information management system on people experiencing homelessness.
Institutional context and objectives
Local authorities in Scotland hold statutory responsibilities to offer settled accommodation to households assessed as unintentionally homeless14 or threatened with homelessness, and to provide temporary accommodation until suitable settled housing becomes available. As part of this duty, authorities make inquiries into applicants’ circumstances and must adhere to statutory guidance, including inter alia recording requirements, as set out in the Code of Guidance on Homelessness, introduced in 2005 and last updated in 2019 (Scottish Government, 2019[23]).
The underlying statutory framework developed over time (Scottish Government, 2025[24]). The duty to assist homeless households was first introduced through the Housing (Homeless Persons) Act 1977, now consolidated in the Housing (Scotland) Act 1987. The Housing (Scotland) Act 2001 and the Homelessness etc. (Scotland) Act 2003 progressively expanded the population entitled to local authorities’ full statutory duties, culminating in the abolition of the priority need test. The Scottish homelessness statistics series dates back to 2002, coinciding with the 2001 Act’s extension of local authorities’ duties (Scottish Government, 2025[24]).
Local authorities submit the underlying data to the Scottish Government, the lead authority for data collection on a quarterly basis, and the resulting statistics are accredited as Official Statistics by the Office for Statistics Regulation. The statistical output comprises an annual report, which is supported by an update every six months. The annual publication serves as the comprehensive statistical release, providing a detailed overview of an individual’s path through (and out of) the system, including the provision of temporary accommodation and the achievement of settled accommodation outcomes. The six‑monthly update complements this by offering an earlier indication of emerging trends in the same core dimensions ahead of the more complete annual data (Scottish Government, 2025[24]; Scottish Government, 2026[25]).
Data collection approach and methodology
The Homelessness in Scotland bulletin includes information relevant to additional ETHOS Light categories, such as experiences of unsheltered homelessness (ETHOS Light 1) and living situations immediately prior to homelessness (as defined in the Scottish context), including stays with family or friends (ETHOS Light 6) (OECD, 2024[13]).
Data are collected throughout the three‑stage statutory homelessness process: the application stage, the assessment stage, and the outcome stage. As part of this process, applicants are asked whether they experienced unsheltered homelessness either on the night before their application or within the previous three months, allowing the bulletin to report on the ETHOS Light 1 category.
Information is also collected on the circumstances from which households became homeless. These data underpin ETHOS Light 6 reporting, since family or friends remain the most common living situation prior to homelessness, with 26% of households previously living in parental or family homes and 20% having stayed with friends or partners. Correspondingly, “asked to leave” is the most frequently cited main reason for a homelessness application, accounting for around one‑quarter of cases. Among outcomes for unintentionally homeless households at the end of a homelessness support process, 4% moved in with friends or family. Notably, such cases are not considered settled accommodation.15 Next to the homelessness circumstances, extensive demographic information is collected, as well as information on circumstances such as having served in the armed forces, and details on support needs (Scottish Government, 2025[24]).
Strengths and limitations
The use of information management systems with data provided by local authorities allows for detailed, timely and comprehensive data collection focussed on individuals, which for example allows for the identification of repeat homelessness cases or support needs of people experiencing homelessness. The quarterly data flow resulting in an annual statistical publication with biannual updates enables relatively frequent updates, compared to other data collection approaches. As data collection is embedded within routine homelessness applications, additional administrative burden is limited, and the data collection approach can be assumed to be cost-effective.
The annual Homelessness in Scotland report highlights several implementation challenges and limitations of the Homelessness bulletin (Scottish Government, 2025[24]). Quarterly submissions may be affected by IT issues, data entry delays, or quality assurance adjustments. In addition, specific data quality issues have arisen, such as errors linked to management information system updates in specific local authorities and occasional incomplete closure of cases. Flow data can involve potential double‑counting, where individuals make more than one application within a year. Importantly, in line with the limitation of service-based methods in general (OECD, 2025[2]), the approach does not capture households experiencing homelessness who do not engage with local authorities. Some individuals may not self-present because they are not aware of the available support services or do not believe they have access to them. Lastly, the operationalisation of the ETHOS Light 6 category differs substantially from that of other countries, by considering temporary stays with family or friends as a precursor to homelessness, rather than a specific form of homelessness, which limits international comparability.
Lessons from information management systems
Information management systems provide continuous, individual-level data, thereby supporting tailored service delivery. The Scottish example, in which homelessness measurement is based on statutory support provided by local authorities, demonstrates the timely monitoring and profiling enabled by IMS.
However, as with service-based methods, the resulting estimates are limited to individuals who engage with the system and thus do not capture non-users. Furthermore, the purpose of the IMS – for example, in the Scottish case, to provide a detailed overview of how the homelessness support process functions – may influence the definition and operationalisation of homelessness. In Scotland, ETHOS Light 6 is interpreted as the circumstances from which a person became homeless – more specifically, if they were staying with family or friends before applying for homelessness support. Operating an IMS also requires significant technical capacity and clear data governance frameworks, as demonstrated by data quality issues in some Scottish local authorities.
References
[17] Australian Bureau of Statistics (2022), Delivering the 2021 Census, https://www.abs.gov.au/census/about-census/delivering-2021-census#cite-window2 (accessed on 16 February 2026).
[19] Australian Bureau of Statistics (2021), Census data supports essential services for people experiencing homelessness, https://www.abs.gov.au/media-centre/media-releases/census-data-supports-essential-services-people-experiencing-homelessness-1 (accessed on 9 June 2026).
[21] Australian Bureau of Statistics (2021), Estimating Homelessness: Census methodology, https://www.abs.gov.au/methodologies/estimating-homelessness-census-methodology/2021 (accessed on 16 February 2026).
[20] Australian Bureau of Statistics (2012), Methodology for estimating homelessness fromt the Census of Population and Housing, https://www.ausstats.abs.gov.au/ausstats/subscriber.nsf/0/61102520832D348ECA257A6F0012AF91/$File/2049055001_2012.pdf (accessed on 16 February 2026).
[10] Benjaminsen, L. et al. (2020), “Measurement of Homelessness in the Nordic Countries”, European Journal of Homelessness, Vol. 14/3, pp. 149-169, https://www.feantsa.org/files/Observatory/Journals/Volume-14/V14-3/EJH_14-3_A7_web2.pdf.
[8] Bundesministerium für Arbeit und Soziales (2022), Ausmaß und Struktur von Wohnungslosigkeit - Der Wohnungslosenbericht 2022 des Bundesministeriums für Arbeit und Soziales, Bundesministerium für Arbeit und Soziales, https://www.bmas.de/SharedDocs/Downloads/DE/Soziale-Sicherung/wohnungslosenbericht-2022.pdf?__blob=publicationFile&v=4.
[9] Bundesministerium für Wohnen, Stadtentwicklung und Bauwesen (2024), Wohnungslosenbericht der Bundesregierung - Ausmaß und Struktur von Wohnungslosigkeit, Bundesministerium für Wohnen, Stadtentwicklung und Bauwesen, https://www.bmwsb.bund.de/SharedDocs/downloads/DE/veroeffentlichungen/wohnen/wohnungslosenbericht-2024.pdf?__blob=publicationFile&v=1.
[14] Centre for State-Subsidised Housing Construction (2026), Homeless people 2025, https://julkaisut.valtioneuvosto.fi/server/api/core/bitstreams/e0bb13a1-c48d-4f0a-8cb9-53c2c7d9eb9f/content.
[15] Centre for State-Subsidised Housing Construction (2025), Homelessness survey instructions.
[12] Dyb, E. et al. (2021), “Governing Through Definitions and Numbers: Analysis of the Nordic Homeless Registrations as Instruments of Governing Homelessness”, European Journal of Homelessness, Vol. 15/3, pp. 187-201, https://www.feantsa.org/files/Observatory/Journals/Volume-15/V15-3/EJH_15-3_A13.pdf.
[7] Government of Canada (2026), Everyone Counts 2025: Results from the Annual Enumeration of Homelessness, https://housing-infrastructure.canada.ca/homelessness-sans-abri/reports-rapports/everyone-counts-2025-tout-monde-compte-eng.html.
[3] Hermans, K. et al. (2025), European Homelessness Counts Final Report Year 1, https://www.kuleuven.be/lucas/en/images/1-final-report-eu-counts.pdf.
[6] Housing, Infrastructure and Communities (2024), Everyone Counts - A Guide to Point-in-Time Counts in Canada, https://homelessnesslearninghub.ca/wp-content/uploads/2024/04/Guide-to-Point-in-Time-Counts-in-Canada-4th-Edition-2-1.pdf.
[5] Housing, Infrastructure and Communities Canada (2025), Everyone Counts - A Guide to Point-in-Time Enumerations in Canada, https://homelessnesslearninghub.ca/wp-content/uploads/2025/07/Guide-to-Point-in-Time-Enumerations-in-Canada-1st-Edition.pdf.
[22] Král, O. et al. (2025), Sčítání osob z vybraných kategorií klasifikace ETHOS, SocioFactor, s.r.o., https://socialnibydleni.mpsv.cz/download/dokumenty/156-zaverecna-zprava-ethos_verze_6_fin.pdf.
[4] OECD (2026), OECD Country Notes: Data on Homelessness and Housing Exclusion, https://www.oecd.org/en/topics/sub-issues/affordable-housing/homelessness.html.
[2] OECD (2025), OECD Monitoring Framework to Measure Homelessness, OECD Publishing, Paris, https://doi.org/10.1787/3e98455b-en.
[13] OECD (2024), OECD Country Notes on Homelessness Data, https://www.oecd.org/en/topics/sub-issues/affordable-housing/homelessness.html.
[25] Scottish Government (2026), Homelessness in Scotland: update to 30 September 2025, https://www.gov.scot/publications/homelessness-in-scotland-update-to-30-september-2025/pages/introduction/.
[24] Scottish Government (2025), Homelessness in Scotland: 2024-25, Scottish Government, https://www.gov.scot/publications/homelessness-in-scotland-2024-25/.
[23] Scottish Government (2019), Code of Guidance on Homelessness - Guidance on legislation, policies and practices to prevent and resolve homelessness, https://www.gov.scot/publications/code-guidance-homelessness-2/documents/.
[16] Statistics Finland (2023), Report on statistics on homelessness.
[18] Stats NZ (2025), Estimated resident population (2023-base): At 30 June 2023, https://www.stats.govt.nz/information-releases/estimated-resident-population-2023-base-at-30-june-2023/ (accessed on 8 June 2026).
[1] Teller, N. et al. (2025), European Homelessness Counts Handbook, European Commission Directorate General Employment, Social Affairs and Social Inclusion, https://www.kuleuven.be/lucas/en/images/handbook-eu-counts-final.pdf.
[11] VIVE (2024), Hjemløshed i Danmark 2024 - National kortlægning, https://www.vive.dk/da/udgivelser/hjemloeshed-i-danmark-2024-dx3jdedv/.
Notes
Copy link to Notes← 1. In Canada, hidden homelessness includes people who are i) living temporarily with others, but without the guarantee of continued residency or immediate prospects for accessing permanent housing (e.g. couch surfing), corresponding to ETHOS Light 6; ii) people accessing short term, temporary rental accommodations without security of tenure and with no access to a safe and permanent residence (e.g. self-funded motel/hotel) (Housing, Infrastructure and Communities, 2024[6]). A question on hidden homelessness has been added to the Population Census, which is expected to produce the first national estimate of ETHOS Light 6 category in Canada (Government of Canada, 2026[7]). See also the discussion in Box 1.1.
← 2. ETHOS Light 5 is included under ETHOS Light 1 for the national estimates on homelessness.
← 3. The following support system areas are considered: Emergency housing assistance; Income support; Addiction support, Youth support; Work and employment; Support for offender; Migration, immigration and support for refugees; Medicine, health and psychiatry; Support services for women; Other.
← 4. Information provided by the DIFU during the OECD fact-finding mission for this report in May 2026.
← 5. Information provided by the DIFU during the OECD fact-finding mission for this report in May 2026.
← 6. Information in this report is based on the 2024 Mapping of Homelessness, which was the latest edition available at the time of analysis. The 2026 edition has since been published.
← 7. Information provided by the Danish Ministry of Social Affairs upon request.
← 8. The so-called National Client Registration System on shelter use existed before – since 1999 (Benjaminsen et al., 2020[10]).
← 9. The homelessness mapping in Denmark takes the ETHOS Light categorisation as a starting point but adjusts it to the Danish context. Each ETHOS Light group includes one or more sub-categories of living situations. In line with the OECD Country Notes, ETHOS Light groups are considered as included in the data collection approach if at least one of the sub-categories is counted (OECD, 2024[13]; OECD, 2026[4]).
← 10. Information shared with the OECD by Varke upon request.
← 11. Meanwhile, people marginally housed in caravan parks are not classified as homeless.
← 12. The 2021 Census involved some departures from standard procedures due to COVID‑19 movement restrictions.
← 13. Information provided by the Czech Ministry of Labour and Social Affairs.
← 14. In Scotland, homelessness is defined not only as a lack of accommodation in the United Kingdom, but also as having accommodation that cannot reasonably be occupied, for example due to a threat of violence.
← 15. Settled accommodation only includes only local authority tenancies, registered social landlord tenancies, or private rented tenancies.