This week homelessness has been in the headlines with a report published by homelessness charity Crisis predicting stark increases if further action is not taken. Here, Prof Glen Bramley, author of the report, takes us through some of the methodology, and difficulties in modelling homelessness.

The analysis I conducted in a report published by CRISIS this week shows that homelessness has gone up by 33% in the last 5 years (across Great Britain), and presents forecasts which suggest we should expect this to increase further in future (e.g. 26% over next decade, more in the long run). Steeper rises are expected in England, especially London, than in Scotland or Wales, with much steeper rises in some forms of homelessness (such as rough sleeping, and unsuitable temporary accommodation).

Such bold, headline-grabbing statements are valuable and probably justified in getting media, policy and political attention given to a pressing social issue. But they do, with a little reflection, beg some important further questions, certainly from a research perspective:

• What do we mean by homelessness here? Is there a credible, useful, well-understood definition?
• How do we measure or count homelessness, given such a definition? Are current measures credible, useful and fit for purpose?
• How can we forecast future levels of homelessness? What do we know about the key drivers?
• So what could or should government (and/or civil society) do about homelessness? Can policies and practices change these outcomes?

Homelessness is clearly a social problem which attracts widespread concern. In that sense it is similar to unemployment, poverty, and other social ills, which have elaborate and well-developed official measures underpinned by substantial data collection exercises. In the case of homelessness, I would suggest that, while we have some official measures, there is a mis-match between these and the most useful definition(s);a lack of credibility about some measures; and an almost total lack of evidence about some aspects of the phenomenon.

In this research, working with the national charity CRISIS, we have agreed to focus on a measure of ‘core homelessness’, (with analysis on wider homelessness due to be published later this year) which covers what most people would agree represents people with the most acute and immediate problems, i.e. those people we can say are ‘homeless now’. Such a definition can more plausibly claim to command general public support (a stable consensus), while avoiding significant practical problems of double-counting and conceptual problems of mixing ‘stock’ and ‘flow’ measures. The agreed definition includes the following elements:

1. rough sleeping;
2. sleeping in cars, tents, public transport;
3. unlicensed squatting or occupation of non-residential buildings;
4. staying in hostels, refuges and shelters;
5. unsuitable temporary accommodation (e.g. B&B, non-selfcontained, out of area placement); and
6. ‘sofa-surfing’, i.e. staying with non-family, on short term basis, in overcrowded conditions.

It is immediately apparent that for some of these categories ((2), (3)) there is effectively no systematic information; for category (1) there are official ‘counts’ reported but these are completely inconsistent in method/coverage and lacking credibility as an adequate measure, as demonstrated through alternative estimation strategies such as the use of retrospective survey questions. While there are reasonable records of hostel place numbers and occupancy rates (4), unsuitable temporary accommodation (TA, (5)) is only measured currently for those households placed there by local authorities. Sofa surfing (6) can be approximately identified in some household surveys, but the resulting numbers appear somewhat inconsistent and probably understate numbers of single persons in this situation.

Local and national government collect and publish quite a lot of data on homelessness, but this is mainly geared to reporting activity under the statutory framework of the Homeless Persons Act 1977 as amended by subsequent legislation. This legal/administrative focus leads to the collection of detailed information on households who are ‘accepted’ as homeless by the local authority, which until recently in England and Wales meant effectively households who were in ‘priority need’ groups (mainly families, and people vulnerable through age, disability etc.), excluding most single homeless people. This group do not map at all closely onto the preferred definition of core homelessness we used in this study: it is mainly presented as an annual flow rather than a stock measure; it omits the majority of core homeless (who are single) and the large minority who never approach the local authority; and it includes a group who rarely at any stage in the process are actually homeless on our core definition. The statutory homeless in UK are largely families who are threatened with homelessness but who are then accepted and typically placed in adequate temporary accommodation (self-contained social  or private rented units), without ever actually being literally homeless, or not for an extended period of time. In some areas, where pressures are intense or local authorities have not managed the process well, more of this group may be placed in unsuitable temporary accommodation, such as Bed and Breakfast, and they would feature in our core homelessness measure, as they would if their previous situation involved sharing with others in overcrowded conditions.

The official measures, then, are some sort of measure of demand/need, as filtered through formal eligibility rules, but they are not an adequate measure of core homelessness. Reforms entailing more duties towards single homeless including prevention and relief activities have been, or are being applied in different parts of the UK, and these will help to overcome some of the above limitations, although not completely.

If we can’t measure core homelessness very well, what hope do we have of forecasting its future trajectory? Well, I would argue that we can obtain a range of estimates for key numbers, in some cases from multiple sources which enables some degree of triangulation, or specifying of a range. So, taking the obvious example of rough sleeping, in England we can derive a ‘low’ estimate from the local authority counts (supplemented by ‘CHAIN’ in London); our middle estimates are a blend of a former admin system for housing-related support (Supporting People), a specialist survey of ‘multiple exclusion homelessness’, and a retrospective household survey (Poverty and Social Exclusion Survey), while a ‘high’ estimate may be derived from special surveys of ‘Destitution’ carried out for a major national charity.

Further, some of these sources (particularly the large-scale surveys, like PSE, Understanding Society, or Scottish Household Survey) enable statistical modelling of the probability of experiencing homelessness as a function of individual and contextual demographic, socio-economic and housing factors. The relationships captured in these models can then be harnessed to predict future risk of homelessness. We have recently demonstrated the power of such modelling to draw out the striking social differences in such risks. This provides a clear body of evidence about the key drivers of homelessness, with a notable emphasis on poverty, as well as demographics, urban location, housing affordability/pressure and supply, and local policy measures such as prevention. This evidence can then be linked with predicted future values for those driving factors, derived within an overall model of the UK housing system/market at sub-regional level.

The final question posed at the outset was the ‘so what?’ question: So what could or should government (and/or civil society) do about homelessness? This research shows clearly that ever-rising homelessness is not inevitable and can be reversed. Concerted government policy action (allied to activities of voluntary sector) can turn the tide and reduce homelessness (as shown in Scotland, where numbers have fallen). Whilst a challenge to completely eliminate homelessness,  the analysis shows through scenarios that certain policies could change the outcomes significantly; relevant policies include. reversing certain welfare reforms/cuts, increasing prevention, increasing housing supply, and correcting regional imbalance.