Measuring homelessness: The Paris street count
By Marissa Plouin (Marissa.PLOUIN@oecd.org) and Ali Bargu (Ali.BARGU@OECD.org), Directorate for Employment, Labour, and Social Affairs (OECD)
Late into a recent evening in January, the city of Paris, together with 30 neighbouring municipalities, mobilised around two thousand volunteers to take stock of the excluded and uncounted people without shelter, as part of its sixth annual street count; la Nuit de la Solidarité.
Street counts like this one in Paris are a common approach to collect data on how many people in a specific geographic area, at a given point in time, are “sleeping rough” – that is, spending the night on the street, in public spaces, or in other places that are not considered appropriate shelter. Similar counts have been undertaken in cities across the OECD, including Berlin, Los Angeles, Sydney, Toronto and Bogotá.
Such exercises are an invaluable tool to provide statisticians and policy makers with an indication about the extent of homelessness now and over time. For the past six years, Paris’ street count has provided a regular snapshot of people without shelter in the capital city. Preliminary results from the 2023 count reveal a 16% increase in rough sleeping relative to 2022. Meanwhile, data on homelessness have been harder to come by at the national scale. The last official government estimate of homelessness in France dates back to 2012, and the Fondation Abbé Pierre recently estimated that the number has more than doubled over the past decade. Nevertheless, there are limits to what street counts can and cannot tell us about homelessness.
How does the street count work in Paris?
Starting around 10pm, small teams of volunteers walked every street in one of 355 geographic sectors. For every individual, family and group encountered, they completed a questionnaire (like this one) with basic observable information. For those willing to be interviewed, the volunteers could ask more targeted questions – with responses to remain strictly anonymous – about the person’s background and access to housing and social services (e.g. “Where do you intend to spend the night?” “How long have you been without shelter?” “Where do you eat meals, shower or find other types of support?”). Once the team covered every street in the assigned sector, they returned completed questionnaires to the neighbourhood headquarters for an initial quality control. Questionnaires were subsequently transferred to the city’s urban planning agency for more detailed analysis, with results to be summarised in an annual report, publicly available on the city’s website (see the 2022 edition, in French).
Why are street counts important?
Street counts can help estimate the extent of one of the most visible forms of homelessness – sleeping rough – at a given point in time. When they are conducted regularly and according to a consistent method, they enable policy makers to monitor rough sleeping trends over time, in order to better plan and target public support. For instance, data from past counts in Paris led to the opening of storage spaces for people sleeping rough, as well as specific housing support for women, who represented about 10% of rough sleepers in 2022 and again in 2023. Street counts may also provide some (albeit often limited) indication of the profiles and needs of rough sleepers. In some cases, street counts may help shift public perceptions around homelessness, particularly when they involve volunteer enumerators, as in Paris, Los Angeles and Sydney. However, this is not universally the case, with some street counts (e.g., in Belgium) relying instead on experts to carry out data collection.
Why are street counts imperfect?
Nevertheless, street counts are, for many reasons, an imperfect tool.
First, street counts provide a partial picture of the experience of homelessness. They exclude people who are “couch surfing,” living temporarily in motels, or sleeping in non-conventional dwellings (like caravans). Second, street count data are static– at a specific point in time –whereas homelessness is most often a dynamic phenomenon. Many people transition in and out of homelessness over several days, weeks or years. Third, street count data are, in many cases, relatively superficial. Even if some street counts aim to collect detailed information (for instance, Paris has differentiated questionnaires for individuals, families and groups) – on the whole, information is less rich and comprehensive than administrative or registry data. This is in part because some rough sleepers are already asleep, choose not to engage, or otherwise cannot be interviewed (e.g. due to language barriers).
Practical implementation challenges associated with street counts can also affect the accuracy and richness of the data. Given the transitory nature of rough sleepers during the day, the counts generally take place at night: but if you start too early, people may not have settled in for the night and you’ll miss them; if you start too late, many people are sleeping, and data collection will be limited. Street counts are also resource-intensive, relying on many enumerators (around 2000 for the Paris count alone), who require training and supplies. Street counts also depend on relatively favourable external conditions: extreme weather or a global pandemic, for instance, complicates data collection efforts and can limit comparability across time. Due to the COVID-19 pandemic, for example, Chicago adjusted its methodology in 2021 – conducting the count in sample neighbourhoods, rather than city-wide, reducing the scope of information collected, and conducting the count across several days and evenings – limiting comparability with past counts. Other jurisdictions cancelled street counts in 2021 altogether.
What methods can complement street counts?
In many cases, street counts are conducted in coordination with a count of people staying in emergency accommodation, such as night shelters, on the same night, to provide a more comprehensive measure of the number of homeless people. In 2022, the Paris street count was, for the first time, conducted in parallel to the five-year census of “mobile habitats and homelessness” organised by the national statistics office, INSEE.
Complementary data collection approaches can help complete the picture, improving our understanding of the prevalence of homelessness, the profiles of people experiencing homelessness, and the many different pathways into housing precarity. For instance, service-based surveys collect data from service providers that typically support people experiencing homelessness – like soup kitchens, food banks, drop-in health centres, shelters and services for victims of domestic violence. Such surveys can be more effective than street counts in accounting for homeless women, young people and LGBTI+, who are less likely to sleep rough or use shelters. In addition, administrative data that are routinely collected by different organisations (such as health data, criminal justice records, social services data) are used in some countries to extrapolate the number of people experiencing homelessness. By-name lists collect and regularly update identifiable, non-anonymised information on people experiencing homelessness, which is used by case managers to direct individuals to relevant support services. New technologies are being piloted to improve data collection, such as the use of satellite technology with AI by the Office of National Statistics in the U.K.to identify specific features in a landscape (caravan parks, displaced populations), along with mobile applications to conduct counts (for instance in Spain). These are just a few examples of how new technology can help to improve measurement, but they also raise thorny ethical questions.
Measuring homelessness is hard, regardless of the collection method
Homelessness is, by its very nature, hard to define, measure and compare across places. There is no internationally agreed upon definition of what it means to be homeless, and OECD countries do not define it in the same way: 13 countries restrict the definition to people who are sleeping rough, and/or living in shelters or other emergency accommodation, whereas 11 countries also count as homeless people living temporarily in hotels and/or doubled up with friends and family. To facilitate measurement and comparison, FEANTSA and the European Commission developed the ETHOS light typology to provide a common language to assess homelessness, yet the typology is not universally applied in data collection and reporting.
How the OECD is helping governments improve data on homelessness
With support from the European Commission and in the context of the Lisbon Declaration on the European Platform on Combatting Homelessness, the OECD is helping governments to close the measurement gap and develop effective solutions to end homelessness. As part of this work, the OECD is mapping the existing evidence base and data collection methods in OECD and EU countries; developing a monitoring framework to help governments better measure and monitor homelessness; and designing a policy toolkit to provide guidance and good practice to combat homelessness. In particular, the Monitoring Framework will summarise the key features, advantages and limits of data collection approaches, and offer practical advice for combining different approaches to assess homelessness more comprehensively.
Better data are an essential piece of the puzzle towards improving our ability to prevent homelessness and to develop more sustainable pathways out of homelessness.