Official Statistics

Rough sleeping snapshot in England: autumn 2021 - technical report

Updated 31 March 2022

Applies to England

1. Introduction

1.1. This technical report outlines in detail how the Rough sleeping snapshot statistics are collected, the quality assurance processes in place and the limitations of the statistics so users are fully informed when interpreting the statistics release.

1.2. Unlike last year, this years’ Rough sleeping snapshot did not coincide with any restrictions which may have impacted people’s risk of rough sleeping. However, during 2020, the government provided extra support for people sleeping rough during the pandemic, in addition to the support already being provided as part of the Rough Sleeping Initiative (RSI) which were in operation at the time of both this years’ and last years’ snapshot and should be noted when comparing figures with previous year’s as they have provided additional support which has helped to protect thousands of vulnerable people during the pandemic.

1.3. By November 2021, there were nearly 4,300 people in emergency & short-term accommodation who would otherwise have been sleeping rough or were at risk of sleeping rough, and over 40,000 people who had already moved on into longer-term accommodation since the pandemic began.

1.4. The Rough sleeping snapshot outputs include:

• A statistical release which provides information about the estimated number of people sleeping rough on a single night between 1 October and 30 November 2021, demographic details (i.e. age, gender, nationality), details about the approach used and the local agencies consulted as part of the snapshot process.
• Data tables are available alongside this release. These include the number of people sleeping rough, demographic information (i.e. age, gender, nationality), the snapshot approach used and any consultations that took place with local agencies.
• Interactive dashboard which enables users to explore trends in rough sleeping since 2010.
• Infographic that summarises the release.
• Annex A which summarises new management information about the support for people sleeping rough and those at risk of sleeping rough, which we have been collecting throughout the COVID-19 pandemic.

1.5. These statistics are designated as Official Statistics. Official statistics are produced to high professional standards as set out in the Code of Practice for Statistics.

1.6. All local authorities across England take an autumn snapshot of people sleeping rough using one of the following approaches:

• A count-based estimate which is a physical count of the number of people seen sleeping rough in a local authority on a ‘typical’ night^{[footnote 1]}.

• An evidence-based estimate is an evidence-based assessment by local agencies, leading to a single snapshot figure that represents the number of people thought to be sleeping rough in a local authority on a ‘typical’ night.

• An evidence-based estimate including a spotlight count, includes an evidence-based assessment with local agencies as well as a street count, which might not be as extensive as the count-based estimate but has taken place after midnight on the ‘typical’ night.

1.7. The snapshot approach has been in place since 2010. From 1998 to 2010, only local authorities with a known or suspected rough sleeping problem (i.e. 10 people or more estimated to be sleeping rough) were asked to conduct a street count. This amounted to about a quarter of all local authorities in England.

1.8. The statistics from 2010 onwards provide a way of estimating the number of people sleeping rough across England on a single night and assessing change over time. The snapshot approach was developed after consultation with local authorities and the voluntary sector to improve the accuracy of recording the number of people sleeping rough in England.

1.9. Since 2010, all snapshot data submitted by local authorities has been independently verified by Homeless Link to ensure the figures are reliable. Homeless Link are the national membership charity for organisations working directly with people who become homeless in England.

1.10. The changes that were made to the approach and the definition of people sleeping rough mean that figures produced after 2010 can’t be directly compared with previous statistics.

1.11. DLUHC published a technical note explaining the differences between the current snapshot statistics and the previous published figures.

1.12. See DLUHC’s previous rough sleeping statistics.

1.13. Since 2016, the snapshot also included demographic information about people found sleeping rough (i.e. age, gender, nationality). Demographic data which used the same categories became available from 2017^{[footnote 2]}.

1.14. Demographic information was collected either by asking the person found sleeping rough or by referring to existing knowledge (if that person was already known). If a person refused to give, or did not want to share, their personal information, they would still be recorded as sleeping rough and demographic information including age, gender and nationality would be recorded as ‘Not known’, if unknown. If it was not possible to record demographic information, due to an individual sleeping for example, then it was recorded as ‘Not known’.

1.15. The title: ‘Rough sleeping snapshot in England: autumn 2021’ was used to make clear that these statistics provide information on the number of people sleeping rough on a single night in autumn, rather than the total number of people sleeping rough across the year.

1.16. The language used in both the guidance for local authorities and statistics publication makes clear that all snapshot approaches were a form of estimate of the number of people sleeping rough. Consequently, all snapshot approaches record only people seen, or thought to be, sleeping rough on a single ‘typical’ night. The figures do not include everyone in a local authority with a history of sleeping rough, or everyone sleeping rough in a local authority across the October-November period.

1.17. Homeless Link have provided further guidance to local authorities about changing their approach for measuring rough sleeping from year to year to ensure the approach chosen provides the most accurate estimate and that the reasons for changing approach were valid. Additional questions about the approach chosen by a local authority have been added to the verification process to ensure closer scrutiny of any approach changes that may have been made by local authorities.

1.18. This statistical release followed a Reproducible Analytical Pipeline so that a transparent audit trail of the analysis could be followed.

2. Recent improvements

2.1. The Office for Statistics Regulation (OSR), which is the regulatory arm of the UK Statistics Authority recently completed a compliance check of the Rough sleeping snapshot statistics against the Code of Practice for Statistics. This review considered the value, quality and trustworthiness of these statistics in relation to the Code.

2.2. In response to the areas of improvement that OSR identified in the compliance review, further changes we have made this for this year’s release include:

• Suppressing small values between 1 and 4 and rounded national and regional totals to avoid any potential for disclosure. The demographic tables are also only displayed for regional and national breakdowns. This has been done for historic figures as well for consistency purposes.

• Publishing for the first time a detailed commentary around the new management information about the support for people sleeping rough and those at risk of sleeping rough, which we have been collecting throughout the COVID-19 pandemic, alongside the annual rough sleeping snapshot statistics. This management information is published in Annex A.

3. Related statistics

3.1. The Rough sleeping snapshot does not provide a definitive number of people or households affected by homelessness in England. The term ‘homelessness’ is much broader than people sleeping rough.

3.2. It is difficult to provide an accurate estimate of all homelessness across England. Data used to compile any estimate is collated from different datasets, which sample different subsets of the population over different time frames. Any estimate of homelessness in England will collate datasets that are not discrete from one another, which means some individuals may have been included more than once in the estimated total.

Management information about rough sleeping and the support for people sleeping rough in England

3.3. DLUHC also collect and publish management information about the support for rough sleepers and those at risk of rough sleeping, as part of the ‘Everyone In’ support in response to the COVID-19 pandemic. This management information provides more timely information about people sleeping rough but uses a similar collection approach to annual rough sleeping statistics, albeit without the greater verification of the annual statistics. Alongside this year’s snapshot statistics, we are also publishing for the first time a detailed commentary on the management information about rough sleeping and the support for people sleeping rough in England Annex A . To assess the quality of new management information, we compared the monthly rough sleeping management information estimates with the 2021 Rough sleeping snapshot data (See Figure 3.1). These two different sources were correlated which provides confidence that local authorities are consistent when estimating the number of people sleeping rough on a single night.

Figure 3.1: Scatterplot of Rough sleeping snapshot (autumn 2021) estimates compared to monthly rough sleeping management information estimates (November 2021)

The Spearman (ρ 0.88) and Pearson (r 0.97) correlation coefficients show that the data are positively correlated

Rough sleeping questionnaire

3.4. DLUHC published the initial report on the rough sleeping questionnaire (RSQ), which represents one of the largest survey data collections on people who sleep rough ever attempted in the UK. The report provides insights into the experiences of people who have slept rough including: their support needs, vulnerabilities, experiences of homelessness, use of public services and an estimate of the fiscal cost associated with someone who sleeps rough.

Statutory homelessness

3.5. DLUHC also collect and publish quarterly statistics on the broad characteristics and circumstances of households owed a statutory homelessness duty. This includes households who approach the council for help with housing who are sleeping rough. The Statutory Homelessness statistics includes some but not all of the people provided with accommodation in response to COVID-19 through the ‘Everyone In’ support. Not all individuals assisted through ‘Everyone In’ would have been owed a homelessness duty. Any individual not owed a homelessness duty would not be part of the local authority case level data submission to DLUHC and therefore would not be included in the Statutory Homelessness statistics. People sleeping rough will only be recorded in these statistics if they have made an application that has been accepted by the local authority.

Hidden homelessness

3.6. DLUHC produce other statistics releases that can help build up the wider homelessness picture.

3.7. The English Housing Survey publishes data on the number of people sofa surfing and concealed households in England, which are additional adults in a household who wanted to rent or buy but could not afford to do so.

3.8. The Continuous Recording (CORE) social housing lettings collection publishes data on those moving from homelessness into Local authority/ Private Registered Providers accommodation.

Expenditure on homelessness

3.9. DLUHC publish statistics on local authority revenue expenditure and financing in England. The RO4 return within the Revenue Outturn suite relates to housing services and includes information on local authorities expenditure on homelessness activities.

Deaths of homeless people

3.10. The Office for National Statistics (ONS) publishes experimental statistics on the number of deaths of homeless people in England and Wales. ONS also publishes publishes statistics on the number of identified deaths involving the coronavirus (COVID-19) of people who were homeless.

Hostels and support services

3.11. Homeless Link publishes an Annual Review of Single Homelessness support, which includes information about the number of hostels and shelters in England and available bed spaces, as well as the support and services available to single people who become homeless in England.

Combined Homelessness and Information Network

3.12. The Combined Homelessness and Information Network (CHAIN) is a multi-agency database that records information about people seen rough sleeping by outreach teams in London. It is managed by St Mungo’s, a London-based homeless charity, and is used by organisations working with people sleeping rough in London. Information is recorded onto the CHAIN database by people who work directly with people sleeping rough in London (e.g. workers in outreach teams, day centres, hostels and resettlement teams). CHAIN, however, does not cover ‘hidden homeless’ groups, such as those who are squatting or staying in inaccessible locations to outreach workers.

3.13. The CHAIN data most similar to the single night snapshot provides a total count of all individuals seen sleeping rough by outreach teams in London on at least one night during the October to December 2021 quarter (CHAIN data is published quarterly and annually). Whereas the DLUHC single night snapshot for autumn 2021 shows the number of people seen or thought to be sleeping rough on a single night between 1 October and 30 November 2021.

3.14. Although the CHAIN data is not directly comparable to the single night snapshot, a relationship between the two sources of evidence is not unexpected as they are measuring the same group of people, albeit in different ways.

3.15. As part of our quality assurance checks, we looked at the correlation between the latest Rough sleeping snapshot estimates for all local authorities in London and the latest CHAIN quarterly estimates of rough sleeping in London. A positive correlation was observed between the two sources of data (See Figure 3.2). This provides further confidence in the quality of the Rough sleeping snapshot estimates as they are similar to another independent source.

Figure 3.2: Scatterplot of Rough sleeping snapshot (Oct-Nov 2021) data compared to CHAIN (Oct-Dec 2021) data

The Spearman (ρ 0.81) and the Pearson (r 0.94) correlation coefficients show that the data are positively correlated.

3.16. The [CHAIN database]https://www.mungos.org/work-with-us/chain/) also records more detailed demographic information about people sleeping rough. See the latest CHAIN reports.

3.17. Although the CHAIN database provides more information about the flows and characteristics of people sleeping rough in London, DLUHC’s Rough sleeping snapshot provides a way of estimating the number of people sleeping rough on a single night and measuring change over time. CHAIN data quality is influenced by the number of outreach workers active on any one night. A local authority which does not have a commissioned outreach team may appear to have fewer people sleeping rough than an adjacent borough which does have an active outreach service.

Comparability across the UK

3.18. The devolved administrations publish their own statistics on homelessness. Details of their releases which contain information on rough sleeping are provided below.

3.19. The Scottish Government publishes figures on the number of households applying to the local authority for assistance under homelessness legislation who say they have slept rough the previous night or have reported their housing situation as ‘long term roofless’.

3.20. The Welsh Government publish a national annual rough sleeping monitoring exercise, which includes a two week information gathering exercise followed by a one night snapshot count. This is carried out by local authorities, in partnership with other local agencies to gauge the extent of rough sleeping across Wales. This has not been carried out for the last 2 years due to the pandemic. The Welsh Government is also collecting and publishing management information on persons placed into temporary accommodation and rough sleepers in response to the pandemic.

3.21. The governing legislation for homelessness in Northern Ireland is the Housing (Northern Ireland) Order 1988 (as amended). Unlike the other three UK nations, housing is allocated by the Northern Ireland Housing Executive (NIHE), which covers the whole of Northern Ireland, rather than by local authorities. In Northern Ireland, statistics on homelessness are obtained from the NIHE.

3.22. The NIHE carry out an annual street count in Belfast in partnership with other local agencies and homeless organisations.

3.23. The figures are not directly comparable between countries as they have different approaches, coverage and are carried out at different time periods.

3.24. ONS recently published a report, UK homelessness: 2005 to 2018, and an interactive tool on the comparability and coherence of existing UK government data sources on homelessness. ONS have also published a report on rough sleeping in the UK.

4. Quality assessment

4.1. The UK Statistics Authority (UKSA) published an assessment report covering the Department’s homelessness and rough sleeping statistics in December 2015. See this report and related correspondence between the Department’s Statistics Head of Profession and the Authority’s Director General for Regulation.

4.2. In 2015, the UK Statistics Authority (UKSA) published a regulatory standard for the quality assurance of administrative data. We have used this standard to assess the quality of the Rough sleeping snapshot.

4.3. The standard is supported with an Administrative Data Quality Assurance Toolkit (PDF, 243KB) which provides useful guidance on the practices that can be adopted to assure users about the overall quality of the Rough sleeping snapshot. See Figure 4.1 for a full outline of the statistical production process.

4.4. Overall, the Rough sleeping snapshot has been assessed as A2: Medium Risk. This is because it is considered to have a medium public interest profile, as there is mainstream media interest with moderate economic and/or political sensitivity. The data is also regarded as medium risk in terms of data quality concerns, due to the large range of data suppliers, which includes all local authorities across England. Homeless Link were not able to verify all Rough sleeping snapshots this year. This was because some local authorities did not involve non-commissioned independent local partners or did not follow the guidance on how to conduct a snapshot properly.

Table 4.1: Assessment of the level of risk based on the Quality Assurance Toolkit

Risk/Profile Matrix Statistical Series	Administrative Source	Data quality concern	Public interest	Matrix classification
Rough sleeping snapshot	Local authority rough sleeping counts and estimates	Medium	Medium	Medium Risk (A2)

Figure 4.1: Rough Sleeping Quality Assurance Process Flowchart An outline of the statistical production process and quality assurance carried out to produce this year’s Rough sleeping snapshot

5. Quality assurance during collection

5.1. Homeless Link have been funded by DLUHC since 2010 to provide guidance to local authorities who conduct the Rough sleeping snapshot and to independently verify all local authorities’ snapshot estimates. Homeless Link’s role includes:

• Publishing a toolkit that supports local authorities to produce an accurate snapshot, whether by a count-based estimate or an evidence-based estimate.
• Providing support to local authorities in planning and carrying out the data collection.
• Verifying each Rough sleeping snapshot.
• Provide training and support to verifiers.
• Cross-checking the data submitted by local authorities to DLUHC with the numbers Homeless Link has verified for each local authority (i.e. the single night estimate of rough sleeping and demographic information).

5.2. The Homeless Link toolkit sets out the process that local authorities, and their local agencies, use to arrive at their snapshot estimate of people sleeping rough on a single night in autumn. Key documents include:

• Rough Sleeping Estimates Toolkit
• Rough Sleeping Estimate 2021 Quick Guide
• Forms and resources for local authorities
• Rough Sleeping Count Form
• Count Verification Form
• Estimate Verification Form
• Resources for verifiers
• Lockdown FAQs
• Webinar slides explaining the process

5.3. Local authorities appoint a coordinator to oversee the process and liaise with Homeless Link’s Project Manager for support and to arrange verification. The coordinator can be someone from the local authority but may also be an external person, such as a rough sleeping coordinator or an outreach team leader.

5.4. All coordinators are required to read the Homeless Link toolkit which sets out the process that local authorities, and their local agencies, should use to carry out their respective snapshots of rough sleeping. The toolkit includes guidance around: defining rough sleeping, choosing between the different approaches, collecting demographic data, how the verification process works, step-by-step guidance, templates, verification forms and quick guides for easy reference.

5.5. Each year Homeless Link verify, in person, all count-based estimates of visible rough sleeping on the night. They also aim to attend and verify at least 10% of evidence-based estimate meetings. The remaining snapshots are verified remotely by Homeless Link’s project team.

5.6. Homeless Link uses its own staff, associates and volunteers to verify the Rough sleeping snapshot estimates. These verifiers receive Homeless Link training and must be independent of the local authority they are verifying. The reasons why this independent verification role is important are that it:

• Provides independent oversight of the snapshot process
• Ensures that there is a comparable process, year by year
• Checks that each local authority is using the correct process
• Enables action if local authorities are using the wrong process
• Arrives at an accurate single figure and collates demographic data
• Helps to answer questions and resolve issues
• Asks questions and gathers information about rough sleeping causes and responses in different areas
• Gains feedback to improve next year’s processes

5.7. Verifiers attending count-based estimates observe and check that people counting have followed the published guidance. They will join a counting team and will be back at the base before the other count teams arrive to speak to each counting team. Verifiers are required to keep a tally of numbers and demographics and ask questions of each count team such as:

• Did they make any decisions not to count anywhere (e.g. due to safety concerns)?
• Were there any issues identifying people sleeping rough?
• Were there any incidents or activities that might have changed patterns of rough sleeping?
• Were the safety guidelines followed?
• Was the estimate what they were expecting?

5.8. Verifiers attending evidence-based estimate meetings observe the meeting and complete a verification form. Their role is to be objective and ensure the process follows the published guidance; making sure evidence has been examined and discussed to provide a reliable estimate of the number of people sleeping rough on the night chosen for the estimate. Verifiers at evidence-based estimate meetings should:

• Confirm the coordinator has contacted a range of local agencies, including relevant voluntary and community organisations, and invited them to be part of the estimate process
• Confirm agencies have been given sufficient information to participate in the process and provide the information required
• Confirm a ‘typical’ night has been chosen and that all agencies involved understand the principle of the ‘typical’ night, the definition of rough sleeping and that there is a data sharing protocol in place
• Assess whether enough data and evidence has been gathered to resolve any conflicting information and that double counting has been avoided
• Check if there are any issues or dynamics between local agencies that could influence the meeting
• Observe the meeting and make notes in the verification form
• Listen to the evidence presented for each person thought to be sleeping rough
• Provide guidance or make decisions about specific cases if the group are unsure or disagree
• Make sure demographic data for each person sleeping rough is provided and any gaps in the data are explained and recorded as ‘Not known’
• Ask questions to remind people of the rough sleeping definition and to ensure that decisions are led by evidence

5.9. Verifiers attending count-based estimates and evidence-based estimate meetings in person have to complete a verification form which asks a series of questions about the process. For example, checking that the guidance had been followed, that relevant local agencies were involved, that a single ‘typical’ night was used, and that there was an understanding of the rough sleeping definition. These verification forms have to be returned to the Homeless Link project manager who will review them and then let the local authority know they can submit their snapshot estimate and related demographic information via DLUHC’s DELTA online data collection system. The project manager will note any concerns or recommendations relating to the process which will be reported back to the local authority and also kept by Homeless Link to help inform their risk assessment of other local authorities.

5.10. There are a number of reasons why Homeless Link might not be able to verify an estimate. For example, if there were no independent local agencies involved or there was a lack of reliable evidence that people included were sleeping rough on that night or evidence was not included about known sleep sites or individuals sleeping rough.

5.11. The involvement of local agencies is an essential element of the Rough sleeping snapshot. It ensures transparency about how the snapshot was carried out and enables Homeless Link to speak to independent local agencies as part of the verification process. This year, 38 local authorities (12%) provided figures that were unverified by Homeless Link. This was because they either they did not involve independent, non-commissioned local partners or they did not follow the guidance on how to conduct a snapshot properly. These snapshots are still included in this year’s publication but consideration about the lack of independence needs to be considered when interpreting these figures for these specific local authorities. These were: Birmingham, Bournemouth, Christchurch and Poole, Brent, Cheshire East, Cotswold, Dacorum, Darlington, Dartford, East Hampshire, Harborough, Harrow, Havant, Horsham, Hyndburn, Leicester, North Warwickshire, Nuneaton and Bedworth, Peterborough, Redcar and Cleveland, Rochford, Rutland, Ryedale, South Cambridgeshire, South Gloucestershire, South Holland, South Kesteven, South Staffordshire, Southampton, St Albans, Sutton, Swindon, Test Valley, Tewkesbury, Trafford, Warrington, Welwyn Hatfield, West Lindsey and Wokingham.

5.12. Local authorities may not have involved independent local partners for a number of reasons, including last minute cancellations, self-isolation so they could not be involved and difficulty finding independent partners who were not commissioned by local authority in some way or another to support people sleeping rough as services have been expanded. In some other cases, local authorities involved independent, non-commissioned local partners in the planning of the process but not the actual count-based estimate or evidence-based estimate meeting, so these were unverified.

5.13. Any snapshots not verified in person were validated by telephone. In each case, the telephone verifier will ask the coordinator a series of questions about how they conducted their snapshot to complete a verification form. Once completed, they can confirm that the information be recorded on DELTA.

5.14. In 2021, Homeless Link verified 37 of the 39 (13%) count-based estimates in person, liaising with the count coordinator to check that the guidance had been followed in the planning of the count. In each case, the verifier discussed the estimate and demographic information with the coordinator and completed a verification form that was sent to Homeless Link for review. Verifiers were trained by Homeless Link and were not employed by the local authority or its commissioned services. For those 2 count-based estimates that could not be verified in person, Homeless Link verified remotely following the snapshot and also contacted partner agencies including an independent partner separately to complete verification.

5.15. Homeless Link also conducted a detailed verification of 38 (12%) of the rough sleeping evidence-based estimate meetings, by being present at the online estimate meeting. These were Barnet, Birmingham, Brent, Chelmsford, Cheltenham, Cherwell, City of London, Cotswolds, Elmbridge, Forest of Dean, Gloucester, Haringey, Harlow, Havering, Kirklees, Luton, Melton, Mendip, Merton, Newcastle, North Northamptonshire, North Warwickshire, Peterborough, Redcar and Cleveland, Richmondshire, South Oxfordshire, Scarborough, Sefton, South Holland, Stroud, Teignbridge, Tewkesbury, Thurrock, Tower Hamlets, Vale of White Horse, West Northamptonshire, Wiltshire and Worcester. Homeless Link chose these local authorities at random and via a risk assessment to ascertain whether these local authorities would need direct support with planning or verification of their estimate was carried out. High risk authorities could be those with: high numbers of people sleeping rough, issues with local agencies disputing estimates, concerns about the correct approach being used in previous years, or requests for additional support.

5.16. The remaining 232 (75%) local authority estimates were not selected for verification and underwent a validation process by telephone.

6. Quality assurance prior to publication

6.1. Since 2017, local authorities have been required to submit their annual single night snapshot data via DLUHC’s DELTA online data collection system.

What information is collected via DELTA?

6.2. The Rough sleeping snapshot DELTA form asks local authorities to provide the following information:

• Type of approach used
• Date of count (i.e. ‘typical’ night)
• Whether the approach used was same/different to last year
• Whether there were any activities, incidents or circumstances that affected the snapshot
• Which local agencies were involved
• Total snapshot figures
• Demographic details (i.e. age, gender, nationality)

6.3. The DELTA online data collection system includes in-built rules and validation checks to ensure the data provided is complete and accurate. For example, any large changes compared to the previous year would trigger an alert to the authority and require a text description to explain the reason for the change, in order to submit the information. Specifically, these include a rule to ensure:

• Figures have been verified by Homeless Link before they are submitted via DELTA
• Demographic breakdowns sum to the total snapshot figure provided
• All fields are completed before the form can be submitted
• Date of snapshot was between 1 October 2021 and 30 November 2021
• At least one local agency has been consulted as part of the process before information can be submitted via DELTA

and a Validation check to trigger an explanation from the local authority if the:

• Overall snapshot figure is +/- 20 % or +/- 5 different compared to last year
• Gender breakdown change significantly compared to last year
• Age breakdown change significantly compared to last year
• Nationality breakdown change significantly compared to last year

6.4. All returns submitted by local authorities on DELTA also undergo a final certification by Homeless Link to check that the final figures submitted were the same as those which had already been verified. If there were any discrepancies, these had to be discussed and agreed with Homeless Link before they were certified.

6.5. As well as the in-built validations via DELTA, DLUHC analysts also carry out a series of checks including comparing the latest snapshot figures with other relevant data sources to produce a quality report. These include:

• Outlier checks to see how each overall Rough sleeping snapshot figure compares to previous years.
• Outlier checks to see how demographic figures compare to previous years.
• Comparing London snapshot estimates with CHAIN estimates for October to December.
• Comparing new rough sleeping management information with the Rough sleeping snapshot.

7. Data collection process

7.1. People sleeping rough are defined as^{[footnote 3]}:

People sleeping, about to bed down (sitting on/in or standing next to their bedding) or actually bedded down in the open air (such as on the streets, in tents, doorways, parks, bus shelters or encampments). People in buildings or other places not designed for habitation (such as stairwells, barns, sheds, car parks, cars, derelict boats, stations, or bashes which are makeshift shelters, often comprised of cardboard boxes). Bedded down is taken to mean either lying down or sleeping. About to bed down includes those who are sitting in/on or near a sleeping bag or other bedding.

7.2. The definition does not include people in hostels or shelters, people in campsites or other sites used for recreational purposes or organised protests, those in squats or travellers.

7.3. All local authorities take a snapshot of the number of people sleeping rough in their local authority on a ‘typical’ night which is a single date chosen by the local authority between 1 October and 30 November. The date can change from one year to the next and will not be the same for all local authorities, although there is some coordination across neighbouring local authorities, to avoid double counting.

7.4. All snapshot approaches will not include everyone in the local authority with a history of rough sleeping between October and November but rather record people found on the night of the count or known to be sleeping rough on a single night, evidence of which is provided by local agencies.

7.5. The snapshot takes place in the autumn rather than Summer or Winter where numbers are either likely to be higher, due to warmer temperatures or lower, as more temporary night shelters are made available to ensure people do not sleep rough in very cold weather.

7.6. If there are severe weather conditions on the night of the estimate, the local authority may activate their Severe Weather Emergency Protocol (SWEP), if they have one. This provides support for individuals during extreme weather. If people sleeping rough do take up emergency accommodation as part of the SWEP, they are still included in the Rough sleeping snapshot.

7.7. Local authorities appoint a coordinator to oversee the snapshot process and liaise with Homeless Link’s Project Manager to arrange verification. Homeless Link uses its own staff, associates, and volunteers to verify the Rough sleeping snapshots. These verifiers receive Homeless Link training and must be independent of the local authority they are verifying.

7.8. The local authority starts by arranging a multi-agency meeting with local agencies to bring together the most accurate and up-to-date information on rough sleeping. This enables a decision to be made on the most appropriate approach to be used and the ideal date that the ‘typical’ night for the snapshot should be.

7.9. Local authorities are advised to involve local agencies that will have the most accurate information about who is sleeping rough and may include:

• Outreach teams, hostels, day centres and night shelters
• Housing departments and Housing associations
• Police, community safety teams and park rangers
• Faith groups, soup runs, street pastors and advice agencies
• Social Services, youth workers and probation services
• Health and mental health services
• Refuse collectors, town centre and local retail security/management and street wardens

Count-based estimates

7.10. Local authorities are advised to start planning the count-based estimate of visible rough sleeping at least 6 weeks before the count itself. This gives enough time to carry out a range of key activities including:

• Gathering intelligence from local agencies.
• Preparing a data protection statement.
• Organising volunteers.
• Planning the night’s activities.
• Liaising with the police.
• Arranging a Homeless Link verifier.

7.11. Local authorities are advised that the date of the count should only be shared on a ‘need to know’ basis in case this causes a change in the pattern of people sleeping rough, potentially resulting in a less accurate figure.

7.12. The presence of film crews or photographers on the count is not permitted by the guidance, as this would mean it is no longer a ‘typical’ night. For example, people may move sleep sites to avoid being filmed or photographed.

7.13. Start times should be tailored to meet local circumstances (typically between 12-2am and finish between 3-6am). The earliest permitted start time is midnight. In cities and towns with a busy night time economy a start time of 2am, or even later, is more appropriate to include people who bed down after pubs and clubs close. In rural local authorities, where the majority of people are sleeping rough away from built-up and well-lit areas, counts may be conducted in the hours after dawn.

7.14. There is no fixed duration for a count. It will depend on: the geography of the local authority, the number of counters, the number of people sleeping rough and what occurs as the teams are talking to the people they meet (e.g. getting people to shelter/hubs or dealing with emergencies). In most cases, if appropriate intelligence has been gathered and enough volunteers recruited, counts take no more than three hours to conduct.

7.15. Local authorities without regular outreach teams may do a street needs audit with a smaller team in advance of the count. This helps to gather intelligence about sleep sites so that the count can be planned around known sleeping sites. This is because it is not feasible to check every street/alley/car park in most local authorities and thus, some targeting of teams is needed. However, Local authorities are advised not to do ‘pre-counts’ on the same day as they could disrupt activity. For example, if people get prior knowledge a count is due to take place, they may move elsewhere to avoid being disturbed later in the evening.

7.16. Local authority coordinators are responsible for recruiting volunteer counters to help with the count-based estimate. Many local authorities use volunteers from their own workforce, as well as volunteer counters from the local statutory and voluntary sectors or faith and community groups. For example, police, park rangers, day centre staff, street pastors and advice workers. If there is a local outreach team available to the local authority, they should be fully involved in the count-based estimate.

7.17. Local authorities must involve volunteers who are independent of the council and its outreach team (e.g. non-commissioned) as this strengthens the credibility of the count. Homeless Link will not verify the process if no local agencies, or only those that are part of, or commissioned by, the local authority are involved.

7.18. Local authorities are advised to liaise with neighbouring local authorities and, if possible, organise a count on the same night over a county or wider sub-regional area. This reduces the risk of double counting people who move across local authority boundaries.

7.19. It is unlikely local authorities will have enough volunteer counters to cover every street or park in their local authority. If there are sleep sites that are occupied but inaccessible or unsafe, then local authorities should consider using the evidence-based estimate meeting process including a spotlight count to improve accuracy.

7.20. Local authorities are advised to plan ahead to gain access to non-public areas where people may be bedded down (e.g. hospital or college grounds, parks, building stairwells, car parks). This may include contacting police, park rangers or security firms to gain access to those areas.

7.21. Local authorities should be divided into segments for each counting team. Each counting team will need a map clearly showing the boundaries of their segment and any known areas for people sleeping rough within it. Exact boundaries, identifying individual streets or parts of streets, should be identified and drawn on a map. It is important to be precise about boundaries so that areas are not double counted or missed entirely.

7.22. The areas covered should be well known to the teams and individual counters. The most effective counters are those with knowledge of where people are likely to be sleeping rough in that local authority (e.g. workers from local outreach teams). Therefore, someone with knowledge of local rough sleeping should be included in each team, alongside volunteers from other agencies.

7.23. There should be at least one mobile counting team in a car to cover outlying areas and support other teams if they encounter delays (e.g. providing assistance to a person they find sleeping rough).

7.24. At least one person, usually the coordinator, stays at the base of operations to deal with queries and check in with counting teams during the night. The verifier has to be in the first team to return to the base so that they can speak to each team as they return and ask questions about the information being provided. This ensures that the guidance has been followed correctly.

7.25. The majority of count-based estimates will be verified in person by Homeless Link. A minority may be verified by telephone, in which case we will need to speak with at least three participating counters, including someone independent/non-commissioned (e.g. from a local faith or community group).

7.26. Once the information has been verified, it is submitted to DLUHC where it is quality assured and compiled for publication. For more information about the quality assurance processes involved, see the sections on Quality assurance during collection and Quality assurance prior to publication.

What information is collected on the night of a count-based estimate?

7.27. Homeless Link provide a template data collection form for local authorities to use on the night of the count. This information is used both for the Rough sleeping snapshot, as well as helping the local authority establish plans to reduce rough sleeping. The counters are asked to record the following information, where possible, to help plan local responses for supporting people sleeping rough. This includes:

• Location
• Full name (if known)
• Demographics (i.e. age, gender, nationality)
• Support needs
• Length of time sleeping rough
• Reason why details not known
• Whether had accommodation but not taking it up
• Whether known to outreach
• An action plan to end each rough sleeping for each person found

Evidence-based estimate meetings

7.28. An evidence-based estimate meeting is an effective way to collate the evidence of rough sleeping on the chosen night and to discuss any potential duplicates or misinformation. Where a disagreement arises, or where there are conflicting approaches between agencies on the scale of the issue, a face-to-face meeting is a useful way to agree upon an accurate estimate on the number of people sleeping rough in a local authority.

7.29. Evidence-based estimate meetings rely on agencies being able to share accurate basic information about an individual’s situation on a particular night. Agencies should agree a data sharing protocol with the local authority before the evidence-base meeting takes place. It is important that there is a way to identify individuals, by name or a unique identifier (e.g. initials and date of birth), to minimise the risk of double counting and including people who have been housed and were not sleeping rough on the ‘typical’ night.

7.30. If some agencies decline to join the data sharing agreement, they may still submit information as evidence but more weight will be given to those agencies whose evidence is supported by complete data.

What information is collected to inform the evidence-based estimate meeting?

7.31. Homeless Link provide a data collection form template for local authorities to use as part of the evidence gathering task that is used to inform the evidence-based estimate meeting. This information is used both for the Rough sleeping snapshot as well as a way of helping the local authority establish plans to reduce rough sleeping. The information local agencies are asked to gather includes:

• Location
• Full name (if known)
• Demographics (i.e. age, gender, nationality)
• Evidence that the individual was rough sleeping on the agreed ‘typical’ night
• Other agencies working with an individual

7.32. Agencies should collect evidence for the night of the estimate and bring it to the estimate meeting. If a face-to-face estimate meeting isn’t practical, agencies are asked to submit their evidence in a password encrypted email file or verbally via a telephone conference. The coordinator should collate this evidence and resolve any issues with the relevant agencies.

7.33. If there is already a local forum or network, which brings together key agencies working with people sleeping rough, then this can also be used for the evidence-based estimate meeting (with expanded membership, if needed). As long as sufficient data sharing protocols are in place.

7.34. Each agency should present information on who they think slept rough on the night of the estimate. Evidence should be provided for each person. Examples of evidence that an individual slept rough include:

• Individual seen bedded down by local or partner agency
• Individual known to the agency and it is clear that they do not have accommodation that they can occupy (people sleeping rough over a longer period/with a history of rough sleeping)
• A spotlight count carried out in particular locations on the chosen night
• Individual self-reports sleeping rough over the period alongside evidence that they are sleeping rough (use of services, carrying/storing belongings, appearance - while bearing in mind many people sleeping rough are well-presented, other known factors)
• Information that demonstrates that someone known to be sleeping rough, intermittently, did sleep rough on the night in question (e.g. they weren’t seen in their hostel; were at the day centre early; were discharged from prison or hospital shortly before)

7.35. Spotlight counts are an additional evidence source to include alongside multi-agency intelligence (Spotlight counts should only be used alongside other data sources). This can be particularly effective for local authorities with people sleeping in a mix of urban and rural locations, or where there are fluctuating numbers of people sleeping rough. Spotlight counts must take place after midnight on the ‘typical’ night chosen. This is to avoid counting people who are part of wider street activity but have somewhere to stay and will not bed down for the night. Involving independent agencies will give the spotlight count more credibility but may cause resource issues if the same agencies are being asked to attend the estimate meeting.

7.36. The final decision on the estimated number of people sleeping rough is made by the local authority, taking into account all the evidence and eliminating duplicates or unsubstantiated data. The coordinator might need to collate further information following the estimate meeting in order to resolve any outstanding questions. Particularly where people are known to move between local authority boundaries and where there is a risk of duplication with neighbouring estimates.

7.37. Homeless Link aim to attend at least 10% of estimate meetings as part of the verification process based on a combination of: concern/risk, random allocation, geography and availability. During verification, Homeless Link will check which agencies have been involved. Estimates using the evidence-based process must be able to show more than one data source. Where the process involves no local agencies, or only those that are part of or commissioned by the local authority, Homeless Link will not verify the process.

7.38. After the estimate process ends, the Verifier completes the evidence-based estimate verification form. This is then reviewed by Homeless Link’s Project Manager before being returned to the local authority lead.

7.39. Once the estimate has been verified by Homeless Link, local authorities submit this information to DLUHC. For more information about the quality assurance processes involved, see sections on Quality assurance during collection and on Quality assurance prior to publication.

8. Approach changes

8.1. The number of people sleeping rough overtime was investigated in order to examine the possible effects of local authorities’ changing their approach on rough sleeping estimates. Specifically, we looked at those local authorities who have consistently used the same approach^{[footnote 4]} compared with those who have changed approach^{[footnote 5]}.

Local authorities using the same approach compared with those who have changed approach

8.2. There are several reasons why a local authority may change from an evidence-based estimate to a count-based estimate or vice versa from year to year. For example, a local authority may change from an evidence-based estimate to a count-based estimate if there are a lack of local agencies with intelligence about where people are sleeping rough or if agencies refuse to share their data making it difficult to form a reliable estimate. Conversely, a local authority may change from a count-based estimate to an evidence-based estimate if there are new sleeping sites that are inaccessible or unsafe which would not be included in a count-based estimate.

8.3. Local authorities can change approach from year to year but any decision to do so has to be guided by whether it will produce the most accurate and precise figure. Homeless Link ask for details on the decision-making behind any potential change as part of the verification process to ensure they meet the guidance.

8.4. There were 147 local authorities (48 % of the total) that consistently used the same approach for their snapshot since 2010, compared to 162 (52 % of the total) local authorities that have changed approach at least once since 2010 (See Table 8.1). Local authorities that have changed approach at least once since 2010 have consistently recorded higher numbers of people sleeping rough compared to local authorities that have always used the same snapshot approach. However, this does not take into account the differences in the types of local authorities which have changed approach compared to local authorities that have used the same approach.

Table 8.1: Local authorities who have changed approach or used same approach since 2010

Local authorities that have Changed approach at least once since 2010 relative to local authorities that have used the Same approach since 2010

Approach	Number of local authorities	%
Changed approach	162	52.4
Same approach	147	47.6
Total	309	100.0

8.5. Between 2011 and 2016, the number of people sleeping rough in local authorities that changed approach, compared to local authorities that consistently used the same approach, increased at similar rates. However, between 2016 and 2017, local authorities that had changed approach at least once since 2010 saw an increase in the numbers of people sleeping rough (See Figure 8.1). However, this could be in response to local authorities having identified increased numbers of people sleeping rough and as a result changed approach to more accurately account for this increase. While it is difficult to say exactly what may have caused this divergence between 2016-2019, many of the local authorities that have changed approach are located in London which historically has high numbers of people sleeping rough compared to the rest of England. In 2020, there was a sharp decrease in overall levels of rough sleeping regardless of the approach taken by the local authority. The decrease in the number of people sleeping rough, according to the local authorities involved, was largely due to the ‘Everyone In’ support and other projects funded by the Department, including the RSI.

Figure 8.1: Number of people sleeping rough on a single night: comparing local authorities that have never changed approach relative to local authorities that have changed approach at least once since 2010

8.6. The average (median) year on year difference in the number of people sleeping rough per local authority that have never changed their snapshot approach since 2010, compared to local authorities that have changed their approach at least once since 2010, was investigated (See Figure 8.2). This data would suggest that the average change in the number of people sleeping rough per local authority is broadly similar over time, irrespective of approach changes.

Figure 8.2: Box plot^{[footnote 6]} year on year difference of people sleeping rough: comparing local authorities that have used the same approach since 2010

9. Data quality and limitations

9.1. Accurately estimating the number of people sleeping rough within a local authority is inherently difficult given the hidden nature of rough sleeping. There are a range of factors that can impact on the number of people seen, or thought to be sleeping rough, on any given night.

9.2. Unlike last year, this year’s Rough sleeping snapshot did not coincide with any restrictions which may have impacted people’s risk of rough sleeping. During 2020, the government launched a range of extra support for people sleeping rough during the pandemic, in addition to the support already being provided as part of the Rough Sleeping Initiative (RSI).

9.3. Notwithstanding the pandemic, other factors can affect the number of people who sleep rough on any given night. For example, the availability of night shelters, the weather, where people choose to sleep and the date and time chosen for the snapshot estimate. Whilst local authorities are asked to provide possible reasons for any significant changes in the numbers of people who sleep rough, compared to previous years, the figures in this release are subject to some uncertainty and therefore should be treated as estimates of the number of people sleeping rough on a single night in autumn.

9.4. The single night snapshot in England provides a way of estimating the number of people sleeping rough across local authorities and assessing change over time. There is currently no national mechanism for recording every person who sleeps rough in England across the year. However, the single night snapshot aims to get as accurate a representation of the number of people sleeping rough as possible, while acknowledging that there are some limitations. The snapshot approach used in England aligns with a standard approach used in many parts of the world including Canada, the United States and several other European countries including France, Ireland, Italy and Spain.

9.5. All approaches record only those people seen, or thought to be, sleeping rough on a single ‘typical’ night. They do not include everyone in a local authority with a history of sleeping rough, or everyone sleeping rough in local authorities across the October-November period.

9.6. Many people who sleep rough do so intermittently, moving from hostels or ‘sofa surfing’ to sleeping rough and back. Many people sleeping rough bed down in places out of sight to avoid the risk of being detected, which means accurately capturing the number of people sleeping rough across a local authority on a single night is inherently difficult.

9.7. A range of different factors can influence the outcome of the snapshot estimates including:

• Local authority topography: It is more difficult to carry out a count of people sleeping rough in large rural and coastal areas than in urban areas. For example, the surface area to cover on a single night in a rural area is often much larger with a wider range of places people may sleep rough. For example, woods, barns or farm land, which can be difficult or unsafe to access. Whilst in urban areas it may be easier to carry out a street count, there may still be issues with including everyone sleeping rough. This is because people could be sleeping in sites which are difficult and unsafe to access, such as derelict buildings and building sites. Furthermore, people sleeping rough may move across local authority boundaries regularly which may explain some of the year on year differences observed.

• Weather: The weather of the chosen night for the count or estimate may have an impact on the number of people sleeping rough. Severe weather conditions will force many people who normally sleep rough to use a night shelter or hostel, to ‘sofa surf’, or sleep in locations which are hidden. These people would be excluded from the count or estimate which may alter the detected level of people sleeping rough. In some local authorities, we know that the weather was more severe compared to last year and in others it was milder, which may also have impacted on the overall number.

• Time and day of snapshot: The guidance provided by Homeless Link makes clear that the night chosen for the count or estimate should take care to avoid any unusual local factors which may distort the number, such as: club nights, football matches, festivals, charity sleep-outs, higher levels of police activity, or changes to service opening times. Local authorities are also advised to identify an appropriate day of the work week for the estimate, as opposed to a day of the weekend, where people may bed down later or are less visible. Local authorities are advised that start times should be tailored to meet local circumstances. The earliest permitted start is midnight. In cities and towns with a busy night time economy a start time of 2am, or even later, is more appropriate so that people sleeping rough who bed down after pubs and clubs close are still counted. Counts should be completed by 5am. However, in rural areas, where the majority of people are sleeping rough away from built-up and well-lit areas, they may be conducted in the hours after dawn. Local authorities should take care to follow the same time period used each year as closely as possible, unless there is a good reason not to, such as a later night time economy.

Deciding which approach to use for the snapshot estimate of rough sleeping

9.8. Local authorities, together with local agencies, decide which approach and date to use for their snapshot of people sleeping rough. They are advised to use the approach that will most accurately reflect the number of people sleeping rough in their local authority, thus providing the most accurate estimate. The decision about what approach to use is made in September, so that there is enough time to plan the next steps before the snapshot happens in October or November.

9.9. Homeless Link provide guidance to local authorities on the reasons to choose the relevant snapshot approach. These are as follows:

Reasons to choose a count-based estimate:

• People sleeping rough are in sleep sites that will be visible/accessible at night.
• There are changes in the number, population or location of people sleeping rough where sites are visible. These include areas near transport hubs where transient groups sleeping rough make the figures unpredictable from one night to the next, or areas that regularly see people new to the streets.
• There are difficulties forming an evidence-based estimate on the basis of the information available. For example, where there is a lack of local agency intelligence about where people are sleeping rough or where local agencies refuse to share data.
• There is significant disagreement about the numbers between agencies and where sites are visible/accessible.

Reasons to choose an evidence-based estimate meeting:

• Sleep sites are inaccessible (e.g. in woods or dispersed across rural areas).
• Sleep sites are unsafe to access or are hidden from sight.
• The local authority cannot arrange safe access to known rough sleeping sites (e.g. parks, tower blocks) during a street count.
• There is regular intelligence gathering in place by a number of agencies and they are happy to share their intelligence with the local authority.
• Numbers of people sleeping rough are consistently low, they are already in touch with services, and partner agencies agree this is the case (i.e. there are rarely people who are new or returning to the streets).
• The local authority can gather sufficient and reliable intelligence on people sleeping rough on the ‘typical’ night via partner agencies.
• Partner agencies agree to collect information for an agreed night and share this with the local authority for the purpose of the estimate.

Reasons to choose an evidence-based estimate meeting informed by a spotlight count:

• There is a mix of visible/accessible and hidden/inaccessible locations in the local authority.
• The individuals sleeping rough or overall numbers of people sleeping rough in visible/accessible sites change frequently.
• There are conflicting views from local agencies about which approach is right for the local authority.
• Additional intelligence comes to light on the night of a planned street count about hidden rough sleeping (i.e. the process is expanded from count-based estimate to evidence-based estimate meeting).

9.10. As part of the verification process, Homeless Link asks why the specific approach has been chosen and provide feedback to local authorities where there are concerns that the proposed approach might not produce the most accurate snapshot. If there is any evidence that the chosen approach has missed out people known to the local authority, or included people who are not within the rough sleeping definition, Homeless Link may not be able to verify the snapshot.

9.11. Local authorities can adjust aspects of the process from year to year to achieve greater accuracy. Some of the small changes a local authority might make include:

• Choosing the ‘typical’ night earlier (e.g. October) to reduce the risk of severe weather.
• Changing the night of the week in response to a busier night time economy.
• Starting a street count later (e.g. 2am instead of midnight) in response to a busier night time economy.
• Involving new local agencies and/or people with lived experience to improve evidence base/engagement.
• Taking into account a significant change in the local context, such as boundary changes.

9.12. If local authorities change their approach from an estimate meeting to a count-based estimate or vice versa from one year to the next, Homeless Link question the decision-making process to confirm that there is a valid reason to change approach. Areas are advised that undertaking a count at other times of year is not in itself a sufficient reason to change between evidence-based and count-based estimates. Again, if the decision to change approach does not meet the criteria for choosing one approach over another as detailed above, Homeless Link may not verify the snapshot.

10. Uses of the statistics

10.1. The data in this statistical release provides evidence on the prevalence and trends on people sleeping rough in England.

10.2. Within DLUHC they are used:

• For ministerial briefings and correspondence, Parliamentary Questions, Freedom of Information Act cases and to answer public enquiries.
• As background to policy development.
• For monitoring policy, allocating resources, performance monitoring and to support bids for funding from the Treasury.

10.3. Outside of DLUHC users include:

• Local housing authorities for monitoring progress, planning and commissioning services to prevent and tackle rough sleeping.
• Other government departments (e.g. Department for Health and Social Care, Public Health Outcomes Framework).
• The voluntary sector and academics to monitor and evaluate housing policy and for campaigning and fundraising purposes.
• E.U. bodies to compare homelessness across Europe e.g. FEANTSA Overview of Housing Exclusion in Europe.

10.4. Users are encouraged to provide feedback on how these statistics are used and how well they meet user needs. Comments on any issues relating to this statistical release are welcomed and encouraged.

10.5. See the Department’s engagement strategy to meet the needs of statistics users.

10.6. If there are any substantial changes to the collection or publication of the Rough sleeping snapshot then these have to be considered by the Central Local Information Partnership Housing subgroup. CLIP enables the central and local government to work together to deliver an efficient, effective, minimally burdensome information infrastructure for policy development, implementation, service delivery, monitoring and reporting.

11. Revisions policy

The revisions policy has been developed in accordance with the UK Statistics Authority Code of Practice for statistics and the Department for Levelling Up, Housing and Communities Revisions Policy. There are two types of revisions that the policy covers:

Non-scheduled revisions

Where a substantial error has occurred as a result of the compilation, imputation or dissemination process, the statistical release, live tables and other accompanying releases will be updated with a correction notice as soon as is practical. If a local authority notifies DLUHC of an error in the information they have submitted after publication of the release, a decision on whether to revise will be made based upon the impact of any change and the effect it has on the interpretation of the data.

Scheduled revisions

There are no scheduled revisions for this release.

1. A ‘typical’ night should avoid weekends and local events which may impact numbers (e.g. football matches, club nights, festivals, charity sleep-outs, higher levels of police activity or changes to service opening times). ↩

2. Demographic information was collected for the first time in 2016. The totals included ‘Not knowns’ but not all local authorities were able to provide exact numbers of these ‘Not knowns’. Consequently, 2016 demographic data is not directly comparable with subsequent years. ↩

3. This definition has been in place since the snapshot methodology was introduced in 2010. ↩

4. Local authorities that have always used the same approach for estimating the number of people sleeping rough since 2010. ↩

5. Local authorities that have changed their approach for estimating the number of people sleeping rough at least once since 2010. ↩

6. A box plot was used to present the distribution of year on year differences in people sleeping rough in local authorities that have changed their snapshot approach (at least once) against local authorities that have used the same approach since 2010. ↩