The data and research currently presented here is a preliminary collection or relevant material. We will further develop our work on this topic in the future (to cover it in the same detail as for example our entry on World Population Growth).
If you have expertise in this area and would like to contribute, apply here to join us as a researcher.
This entry studies available data and empirical evidence on homelessness, focusing specifically on how it affects people in high-income countries. Homeless people are among the most vulnerable groups in high-income countries.
You can read our entry on Extreme Poverty if you are interested in a broader perspective on economic deprivation and a perspective beyond high-income countries.
All our charts on Homelessness
One of the most common ways to measure homelessness is through so-called ‘point-in-time’ counts of people who are sleeping in shelters or on the streets. These are figures that are intended to reflect the number of people who are homeless ‘on any given night’.
The main source of point-in-time estimates in the US is the Department of Housing and Urban Development, which releases the Annual Homeless Assessment Report to Congress (AHARC). They calculate ‘point-in-time’ estimates by counting homeless people in late January of each year.
The main underlying sources of data used to produce the figures published in the AHARC are (i) registries from shelters and (ii) counts and estimates of sheltered and unsheltered homeless persons provided by care organizations, as part of their applications for government funding.
The counts from the care organizations (called ‘Continuums of Care‘ in the US) come from active counts that are undertaken at the community level, by walking around the streets, using pre-established methodologies.1
In these figures, ‘Sheltered Homelessness’ refers to people who are staying in emergency shelters, transitional housing programs, or safe havens. ‘Unsheltered Homelessness’, on the other hand, refers to people whose primary nighttime residence is a public or private place not designated for, or ordinarily used as, a regular sleeping accommodation for people – for example, the streets, vehicles, or parks.2
In the UK, the government provides official point-in-time estimates of homelessness.
These official estimates come from reports submitted annually by local authorities to the Department for Communities and Local Government. The main sources of information for these reports are either active counts carried out on a single night, or estimates based on information provided by agencies such as outreach workers, the police, the voluntary sector and faith groups who have contact with rough sleepers on the street.
People experiencing homelessness in these figures correspond to people ‘sleeping rough’ and do not take into account most of the sheltered homeless population.
NB. The figures for the UK are not really comparable to those presented for the US above: the definitions and measurement instruments are somewhat different. Below we explore the issue of cross-country comparability across OECD countries in more detail.
The OECD Affordable Housing Database presents a collection of available statistics on homelessness in member countries, in line with definitions used in the corresponding national surveys. Estimates, in most cases, correspond to point-in-time estimates.
In order to introduce some degree of comparability, the OECD provides details regarding what is included in the reported estimates of homeless people, country by country. A full table of definitions of homelessness used for the purpose of data collection in the OECD can be found in Annex 1 of OECD Affordable Housing Database (2016).3
In the visualization we provide an overview of cross-country estimates that are comparable in the sense that they all include exclusively people in the following categories: (i) people living in the streets or public spaces without a shelter that can be defined as living quarters; (ii) people in emergency accommodation with no place of usual residence, who move frequently between various types of accommodation; and (iii) people living in accommodation for the homeless, including homeless hostels, temporary accommodation and other types of shelters for the homeless.
Some issues of comparability remain, since the measurement instruments are not standardized, and data is collected at different points in time (the years for the estimates in the figure range from 2009 to 2015, depending on the country). Nevertheless, these figures give us a rough sense of the relative size of the problem in different countries.
A visualization for other OECD countries reporting estimates for a broader definition of homelessness (countries where counts include additionally people living in institutions and people living temporarily in conventional housing with family and friends), can be found here.
These numbers show that the US is at the higher end of the spectrum together with France, while Southern European countries are at the lower end.
As we have already mentioned, there is no internationally agreed definition of homelessness, and there are no standardized instruments for measuring homelessness – not even among OECD countries. Given this difficulty, some studies have tried to collect information through specially-designed surveys and instruments, such as phone calls enquiring about experiences with homelessness.
Despite obvious limitations with the data (telephone surveys are likely to exclude those who are homeless in the long run), these studies provide estimates that are suitable for comparisons across countries.
The visualization shows the findings from one such study (Toro et al. 2007).4
These estimates come from randomly sampling and interviewing people by telephone in five different high-income countries, asking them about their experiences with homelessness; specifically, whether they had ever had an episode of literal homelessness in their life (where ‘literal homelessness’ means sleeping on the street or in a shelter facility).
As we can see, homelessness is a significant issue in all these countries. In the UK, about 1 out of 13 adults report having slept at least once on the streets or in a shelter in their lifetime.
To the extent that the point-in-time estimates of homelessness tend to be much lower than the lifetime rates of prevalence (a fact that is visible if we compare the estimates for the US and UK in the chart with those discussed above), we can infer that, for the majority of people in these countries, these episodes of homelessness are transitory.
This last point is consistent with the evidence from other studies. Kuhn & Culhane (1998)5 and Benjaminsen and Andrade (2015)6, using data from the US and Denmark respectively, show that the largest group amongst the homeless are the ‘transitionally homeless’, with relatively few and short shelter stays.7
How do the ‘literal homeless’ compare to the ‘extreme poor’ who are not homeless? Researchers have tried to answer this question by matching and comparing groups of individuals who fall under these different categories of vulnerability.
Toro et al. (1995),8 for example, sampled 144 adults in the US, from sites such as soup kitchens that offer shelter facilities for some of those served, as well as shelters that offer food to poor individuals who have housing elsewhere, and constructed three roughly comparable groups: the currently homeless, the previously (but not currently) homeless, and the never-homeless but poor.
Despite the small sample sizes, they find that the never-homeless poor individuals were significantly more likely to be receiving public benefits, were less likely to have a diagnosed mental disorder or problems with substance abuse, and showed lower levels of self-rated psychological distress.
A number of other studies from the US provide similar evidence, suggesting that those who are homeless (in the sense that they are roofless or sleep in shelter facilities) tend to be a particularly vulnerable subgroup of individuals within the poor.9
Of course, the data is far from perfect (small samples, potential sources of bias coming from sampling methods, etc.); but the key message here is that (i) homelessness and extreme poverty in rich countries are closely related, and (ii) we can get a sense of relationship by simply asking.
There is no internationally agreed definition of homelessness. Different governments and organizations use different definitions.
In most countries, different terms are used for different types of situations. The term ‘literally homeless’ is often used to denote the people staying in shelters for the homeless, on the streets, or in other similar settings (e.g., in abandoned buildings, in make-shift structures, in parks). And within the group of people experiencing ‘literal homelessness’, it is common to distinguish between the ‘unsheltered homeless’ and the ‘sheltered homeless’. Unsheltered homelessness is also sometimes referred to as ‘rough sleeping’ or ‘rooflessness’.
Apart from the ‘literally homeless’, there are many other persons who are often classified as ‘precariously housed’. This term is often used to denote people living with a family member or friend for lack of alternatives.
Unfortunately, for the purpose of measurement, estimates are not available across all these groups. In many countries it is common to report together the ‘literal homeless’ with the ‘precariously housed’.
The most common way of measuring homelessness is through so-called ‘point-in-time’ estimates of people who are sleeping in shelters or on the streets. These are figures that are intended to reflect the number of people who are homeless ‘on any given night’.
The main underlying sources of data used to produce point-in-time estimates are (i) registries from shelters (ii) active counts carried out on a single night, or (iii) estimates based on information provided by agencies such as outreach workers, the police, the voluntary sector and faith groups.
Point-in-time estimates are often contrasted to annual – or longer-term – prevalence estimates, typically obtained from surveys asking people about their experiences with homelessness, or from registries of people applying for social housing support.
Understanding the link between poverty and homelessness
It is surprising that we do not have a good idea of how high homelessness rates are among the poor in rich countries.
As we have already mentioned, one way to estimate the prevalence of homelessness is by asking people directly about their experiences with homelessness (e.g. by conducting phone interviews and asking whether people have ever slept rough). This suggests that by incorporating simple questions in income and consumption surveys, it should be straightforward to report official statistics on the prevalence of lifetime homelessness, both for poor and non-poor individuals. Of course, such estimates would be biased because the ‘chronically homeless’ do not appear in traditional surveys. But the numbers would still give us some useful hints regarding the link between poverty and homelessness.
Understanding the link between homelessness and poverty in rich countries is also important because it highlights some of the difficulties that we face when attempting to measure welfare via incomes and consumption.
As Atkinson (2016)10 pointed out in a report for the World Bank, the issue is that the qualitative nature of incomes and consumption, as well as the implicit degree of agency, should be taken into account when measuring poverty. To establish whether the homeless in rich countries are ‘extreme poor’, as measured relative to the International Poverty Line, we need to ask ourselves some difficult questions. Should income from begging, or food from a soup kitchen, be regarded as equal in value to a welfare check? Is money received from selling plasma (as described by Edin and Shaefer 201511) equivalent to a paycheck?
These difficult questions underscore the importance of tracking deprivation across multiple dimensions of well-being, including both standard and non-standard economic indicators. This is our approach at Our World in Data.
Systematic evaluation of policies to reduce homelessness in rich countries
While there is a huge literature in economics studying the effectiveness of poverty alleviation programs in low income countries, there is very little research regarding the effectiveness of policies aimed at ending homelessness in rich countries. In fact, to our knowledge, there are no experimental studies that provide solid evidence regarding policies that causally reduce homelessness.
- Data: Available estimates at national level on the number of people reported by public authorities as homeless. The documentation explains what is included in the number of homeless (i.e. explains the definition used for statistical purposes)
- Geographical coverage: OECD member states
- Time span: Recent years, but not all countries submit estimates that are comparable across time
- Available at: http://www.oecd.org/social/affordable-housing-database.htm
- Data: Point-in-time estimates of sheltered and unsheltered homeless population (using definition of ‘literally homeless’). Estimates are also disaggregated by demographic characteristics.
- Geographical coverage: United States, disaggregated by state
- Time span: 2007-2016
- Available at: https://www.hudexchange.info/programs/hdx/guides/ahar/#reports
- Data: Point-in-time estimates of unsheltered homeless population (referred to as ‘rough sleepers’).
- Geographical coverage: England
- Time span: 2010-2016
- Available at: https://www.gov.uk/government/statistics/rough-sleeping-in-england-autumn-2016
- Reference: Toro, P. A., Tompsett, C. J., Lombardo, S., Philippot, P., Nachtergael, H., Galand, B., … & MacKay, L. (2007). Homelessness in Europe and the United States: A comparison of prevalence and public opinion. Journal of Social Issues, 63(3), 505-524.
- Data: Share of population who have ever been homeless (as per telephone survey results).
- Geographical coverage: Germany, Belgium, Italy, United States, United Kingdom
- Time span: Around 2013 (NB. interviews took place at different points in time within and across countries; but all interviews took place in the period 1999-2003).
- Available at: http://www.uclep.be/wp-content/uploads/pdf/Pub/Toro_JSI_2007.pdf