2049.0.55.001 - Information Paper - Methodology for Estimating Homelessness from the Census of Population and Housing, 2012
Latest ISSUE Released at 11:30 AM (CANBERRA TIME) 05/09/2012 First Issue
|Page tools: Print Page Print All RSS Search this Product|
LIMITATIONS AND ISSUES WITH USING THE CENSUS DATA TO ESTIMATE HOMELESSNESS
Under / over estimation and under-enumeration
Observing homeless people in any data collection is a challenge, and the homeless circumstance may mean that these people are not captured at all in datasets used to count people generally. And not all homeless people will be enumerated in data sets of those homeless people accessing particular services for the homeless. The ABS General Social Survey found that of those who had had an experience of homelessness in the last ten years and who were no longer homeless at the time of interview, only 40% had sought assistance of formal services (Australian Bureau of Statistics, 2012a). While data on people who access services are very important data sources to understanding people who access services, they cannot provide an estimate of the total number of homeless people at one point-in-time. Only the Census offers the opportunity to estimate the people who were likely to have been homeless at that one point-in-time.
However, there is an inherent imprecision in estimating homelessness using the Census of Population and Housing because the Census is not designed to classify people according to whether or not they were homeless on Census night. Variables collected in the Census that were designed for other purposes must be interpreted as proxies for likely homelessness. The ABS methodology includes in its homelessness estimates groups of people who, on balance, were most likely to have been homeless on Census night.
While it may be tempting to overestimate homelessness in some groups to compensate for both under-enumeration and likely under-estimation for some other groups, such an approach would result in estimates of characteristics that did not reflect those of the homeless population, including but not limited to their geographic spread. This may result in the misdirection of policy, funds and services. And while a balance between unavoidable under-estimation and deliberate over-estimation may result, this is unlikely, particularly when there is little information on the magnitude of underestimation. It is also very likely to be very different with each Census, destroying the capacity to monitor change over time. Recognising which groups of homeless people are underestimated in the Census, and using supplementary data sources to understand these groups, will both better address the needs of homeless people, and allow for assessments of change over time in the level of homelessness.
ABS recognises that some groups of people are more likely to be under-enumerated in the Census. Aboriginal and Torres Strait Islander peoples are more likely to be both underenumerated and overrepresented in the homeless population. ABS has developed strategies for each Census aimed at maximising the enumeration of Indigenous persons (Australian Bureau of Statistics, 2012b and 2006).
So called rough sleepers and people staying in supported accommodation for the homeless are also at risk of being under-enumerated in the Census. The ABS develops a homeless enumeration strategy for each Census, and works with homeless service providers to maximise the enumeration of these groups on Census night (Australian Bureau of Statistics, 2012c and 2008).
The ABS Post Enumeration Survey (PES) is used to estimate for the under enumeration of the Australian population in the Census. However it only covers people in private dwellings at the time of the PES, and therefore will not capture homeless people living in non-private dwellings but who were missed on Census night. Also the PES does not capture sufficient information to determine whether a person is homeless. Furthermore, the PES is a sample survey and the likelihood of including a homeless person in a private dwelling is low. For all of these reasons the PES cannot be used to estimate under-enumeration or under-coverage of homelessness estimates derived from the Census.
Key Population Groups
For some key groups, Census variables provide limited opportunity to estimate those likely to be homeless. Three key groups are:
The ABS homeless estimates derived from the Census are likely to underestimate the extent of youth homelessness but there are no data available to determine even approximately the magnitude of the under-estimation.
For some youth (sometimes referred to as 12-18 years or 12-24 years) who are homeless and 'couch surfing', a usual residence may still be reported in the Census. Their homelessness is masked because their characteristics look no different to other youth who are not homeless but are simply visiting on Census night. A usual address may be reported for 'couch surfers' either because the young person doesn't want to disclose to the people they are staying with that they are unable to go home, or the person who fills out the Census form on behalf of the young person staying with them assumes that the youth will return to their home. Homeless youth will be underestimated within the group: 'Persons staying temporarily with other households'.
ABS has not been able to establish any reliable way, with existing data sources, of estimating homelessness among youth staying with other households and for whom a usual address is reported in the Census. Service providers and researchers have indicated that the estimates of homeless youth derived from Census data do not concord with their knowledge about youth homelessness but there is no information to know the extent of any divergence.
Guided by its Homelessness Statistics Reference Group (HSRG), the ABS is continuing to undertake research and development to improve the estimation of homelessness, including youth homelessness. In particular, the ABS has been undertaking a quality study to inform the potential development of a nationally representative homeless school students survey.
Until a robust methodology is developed to measure the level of youth homelessness, ABS will focus on producing transparent, consistent and repeatable estimates that can be used to monitor change over time. Because the ABS methods are transparent, users can assess whether there is any evidence to suggest that the components of homelessness that cannot yet be estimated reliably are likely to be moving differently over time to those elements that can be measured.
For analysis on youth homelessness from the Census see ABS publications: Discussion Paper: Methodological Review of Counting the Homeless, 2006 (cat. no. 2050.0.55.001) and Position Paper - ABS Review of Counting the Homeless Methodology, August 2011 (cat. no. 2050.0.55.002).
Persons displaced due to domestic and family violence
The ABS recognise the difficulties in both enumerating people who are displaced from their home due to domestic and/or family violence, and in classifying all of those that are enumerated correctly as either homeless or not homeless on Census night. Those enumerated in supported accommodation for the homeless will be measured. Some who are in boarding houses, staying temporarily with other households, in improvised dwellings or sleeping rough, or staying in other lodgings such as hotels or motels on Census night and who report no usual address will be classified as homeless. However some will not be able to be distinguished from other people who were visitors on Census night.
Some people who are displaced due to domestic and family violence may not be enumerated in the Census. Out of fear they may not have themselves recorded on a Census form for the dwelling they are staying in. For those who are reported on a Census form as being away from home on Census night, they may be reluctant, for a number of reasons, including stigma, to report having no usual address on Census night. Alternatively, they may have an expectation that they may be able to return to their home in the future and do not see themselves as not having a usual address. As a result they cannot be distinguished from other people who were visiting on Census night and Census based estimates must be recognised as being an underestimate for this group.
The ABS have worked with its HSRG members to look to ways to both improve the enumeration of these homeless people in future Censuses as well as developing alternative sources of information such as the Personal Safety Survey.
Aboriginal and Torres Strait Islander Australians
Aboriginal and Torres Strait Islander Australians have been underenumerated in the Census and estimates of homeless based on Census data will be an underestimation. In the 2006 Census, the under-enumeration of Indigenous Australians was 11.5%. Some of those who were under-enumerated may have been homeless at the time of the Census.
Underestimation of homelessness among those Aboriginal and Torres Strait Islander Australians who were enumerated in the Census may occur as, for some Indigenous people, 'incorrect' information regarding 'usual residence' may have been provided which masks their homelessness.
It is debated in the literature whether the concept of 'no usual address' is appropriate for some Aboriginal and Torres Strait Islander Australians. Morphy (2007) discusses the problems in defining a 'usual resident' and 'visitor' in an Aboriginal and Torres Strait Islander Australian context, as the distinction between 'my country / not my country' is more salient than the distinction between 'resident / visitor'. This issue becomes particularly problematic for people who are highly mobile. Chamberlain and MacKenzie (2008) also discuss the relevance of 'no usual address' to the Indigenous population, as the 'usual address' question is approached with a different cultural frame of reference. They note that it is not culturally appropriate to record 'no usual address' on Census night because 'home' is understood in a different way, particularly when Aboriginal and Torres Strait Islander people are staying with their extended family. Due to the different cultural frame of reference for Aboriginal and Torres Strait Islander people, it is widely assumed that the Western concept of 'no usual address' is under-reported by Aboriginal and Torres Strait Islander Australians (Chamberlain and MacKenzie, 2008). This issue will impact on Census based estimates of homelessness among Aboriginal and Torres Strait Islander persons enumerated in the Census who are classified as being in the categories of 'improvised dwellings, tents, sleepers out', or temporarily visiting friends or relatives while homeless.
While the increased effort in the 2011 Census to improve the enumeration of Indigenous Australians was successful, under enumeration is still large, and the difficulties in differentiating between those who may be homeless remain.
In recognition of the differences in understanding of the concepts of home and homelessness in the Aboriginal and Torres Strait Islander Australian context, the ABS is undertaking further research about how the ABS statistical definition of homelessness may be understood in the Aboriginal and Torres Strait Islander Australian context. The ABS will establish a sub group of the HSRG to provide recommendations to the HSRG for the development of a definition relevant to homelessness in an Aboriginal and Torres Strait Islander context. Any recommendations from this group will be considered by the ABS in both Census enumeration and subsequent estimation of homelessness among Aboriginal and Torres Strait Islander Australians. The ABS will also work to develop a culturally appropriate module on previous experiences of homelessness suitable for inclusion in the National Aboriginal and Torres Strait Islander Social Survey which can be compared to estimates from the total population from the 2014 General Social Survey.
Estimates of Aboriginal and Torres Strait Islander homelessness from the Census should be considered to be an underestimate.
For analysis on Aboriginal and Torres Strait Islander homelessness from the Census see the ABS publications: Discussion Paper: Methodological Review of Counting the Homeless, 2006 (cat. no. 2050.0.55.001) and Position Paper - ABS Review of Counting the Homeless Methodology, August 2011 (cat. no. 2050.0.55.002).
Interpretation of usual residence question in the Census
The 'usual residence' or usual address variable in the Census is designed, for Census purposes, to report on: population characteristics by small area; and to report mobility i.e., current usual address versus address one and five years ago.
Interpretation of Census data on 'usual address' without reference to other information reported in the Census as an approximation to homelessness is incorrect due to the way the question is worded and the intent of the question.
The ABS Census asks people to report a usual address. The instructions for reporting are to write in:
The advice, if enquired, to people completing the Census form and who move around is that a usual address is somewhere you have lived or intend to live for six months or more, otherwise writing 'none' in the usual address question is the appropriate answer.
There are a very wide range of reasons why a person may not have stayed, or be intending to stay, at a particular address for 6 months or more in a particular Census year. In the 2007-08 Survey of Income and Housing about 16% of household reference persons reported having lived at their current private dwelling address for less than 1 year, implying that on average, at least 250,000 people change address each month (Australian Bureau of Statistics, 2009b).
People will have moved from a former usual address for many reasons, for example moving for study or work, or upon retirement. Some of these movers may be temporarily accommodated in their new city or town, and at the time of the Census cannot report a future address of the home that they have not yet rented or bought.
People who moved in July or August, just before the Census, might report their former home as the place they had lived for at least 6 months, but may consider it odd to report this old address as their 'usual' address. It is considered unlikely that people report a former usual address as their current usual address after they have permanently left that address, or left it on a long term basis. The design of the Census 'usual address' question for reporting on mobility, and for supporting population measures, would be undermined if people did report their old usual addresses to which they would not be returning, or not returning for quite some time. By reporting 'no usual address' these people are counted in the population where they are enumerated, and counted as movers from their former usual address.
People moving to step up in either the jobs market or the residential property market, or people capitalising on their lifelong residential investment when they retire, may temporarily not own any property while between investments, but are unlikely to experience the forms of social exclusion that affect people who fit a more traditional view of homelessness. For example, as reported in ABC radio interviews, families moving from Queensland to Karratha to rent a slab on which to park their caravan reported that 'the money was just too good to refuse'. Such families would certainly benefit from cheaper housing options in their new area in the long term, either to rent or to add to their holdings of owned premises, but the issues for social inclusion are less likely to reflect the entrenched disadvantage (or risk of such disadvantage) that characterises the homeless population as defined by the ABS definition of homelessness, see Information Paper: A Statistical Definition of Homelessness (cat. no. 4922.0).
The ABS uses reporting of no usual address as a starting point to identify anyone who may be homeless and then refines this broad inclusion by analysing these people to identify, on balance, groups of people who are unlikely to be homeless. The areas of exploration of those reporting no usual address but on balance as a group are unlikely to be homeless include those identified as overseas students in group houses, 'grey nomads' travelling in their caravans after retirement, and a wide range of other people in visitor only households, especially those renting in holiday destinations, and sharing holiday accommodation with other families that do report a usual residence, and recently arrived migrants or residents returning to Australia who, similar to internal migration situations, have not yet had the opportunity to select their permanent home. Some of these people in these groups are clearly not staying with usual residents in a household, while others are staying in a substantial dwelling that they may own in a holiday destination but which may be a second or third home and is not their 'usual residence' in a Census year.
See Chapter 5 methodology for more information on those people who are, on balance, not classified as homeless.
Identification of non-private dwellings and private dwellings
Both non-private and private dwellings are important in identifying where people who may be homeless are enumerated. NPDs include places such as hotels, motels, staff quarters, boarding houses, prisons, hospitals etc. Once it is established whether people are enumerated in a private or non-private dwelling, different assumptions must be applied to determine whether the occupants in the dwelling are, on balance, most likely to be homeless or not on Census night.
Non-private dwelling are initially identified by a Census Area Supervisor and confirmed as such by the owners and managers of those properties. Census collectors may also find a non-private dwelling missed by the supervisor, and in these cases the usual owner verification is followed. All other dwellings identified as structures which are intended to have people live in them, and which are habitable on Census night, are classified as private dwellings. However, it is possible that some private dwellings may be operated on Census night as an illegal boarding house.
The boarding house classification of NPDs by their owners appears to work reasonably well in field. However, the classification is not designed to only capture boarding houses for the homeless - it also captures dwellings such as regional accommodation serving children from multiple schools, which need to be removed for the purposes of estimating homelessness.
If a private dwelling is being operated illegally on the basis of multiple room by room tenancies, the area supervisor and collector may enumerate it according to its legal and apparent basis of operation as a private dwelling and not as a boarding house. The methodology (see Chapter 5) applies rules for large (apparently) unrelated group households in order to try to identify, and include as homeless accommodation, any private dwellings that may be operating as boarding houses. However this method is limited and may overestimate the number of large (apparently) group households as boarding houses due to the limitations of the Census in capturing all relationships in the household.
The Census only captures the relationships in relation to the first person listed on the household form, and for child relationships to person 2. Therefore, people in the household who are a couple, but neither person has a relationship to person 1, will not be reflected as a couple. Households with 5 or more usual residents may therefore be incorrectly classified as group households and could incorrectly be assumed to be a boarding house for the homeless. When a usual resident on Census night is temporarily absent, failure to take this into account can also lead to wrong assumptions about the nature of both the household and the dwelling.
By just relying on the NPDs classification returned from the field for estimating these populations, it is possible that some supported accommodation dwellings and some illegal boarding houses will be misclassified as private dwellings. The methodology outlined in Chapter 5 shows how these dwellings are treated, in particular to help ensure that people in 'group houses' are treated correctly.
To ensure that people staying in supported accommodation for the homeless are correctly classified, the ABS obtains lists of homeless supported accommodation to correctly flag these dwellings as homeless accommodation. ABS also supplies green stickers to homeless services, who encourage their clients to place the sticker on their Census form and then post their form back to the ABS. The ABS then flags these dwellings as being supported accommodation for the homeless.
Some people living specific living situations in NPDs on Census night are not included in the definition of homelessness. While these living situations lack one or more of the key elements of 'home' identified in the definition of homelessness, the people occupying these places are not regarded as homeless. People in these living situations are not classified as homeless as:
The specific exclusions include:
The definition of homelessness includes residents of boarding houses as homeless if they:
If it is assumed that people living in boarding houses do not have either of these elements and do not have accommodation alternatives, so they are considered homeless.
However, some residents of boarding houses have secure tenure and have access to space and privacy. These people would not be considered to be homeless, even if they do not have accommodation alternatives. The Census has some information which allows the ABS to establish whether a dwelling is most likely to be a boarding house. However it does not offer enough information about occupants security of tenure, or their access to space for social relations. Therefore, determining which occupants are homeless is difficult. In the absence of this information, on balance the ABS has included all people in these dwellings, who are either usual residents (excluding staff) or visitors reporting no usual address, as homeless. This is likely to result in an over estimate of people who are homeless when assessed against the ABS definition. The ABS does, however report those in boarding houses as a separate category to aid policy and service provision. This is because in some State and Territories boarding houses are used to move people out of other forms of homelessness, such as rough sleeping or couch surfing, towards more independent, secure, long-term accommodation.
Although the ABS makes a significant effort to identify boarding houses, both registered and unregistered, the ABS acknowledge that there will potentially be an under estimation of people living in boarding houses in estimates of homelessness, and an overestimation of those living in boarding houses who are likely to be homeless according to the ABS definition. For future Censuses, the ABS will look at ways to improve the identification of boarding houses such as obtaining lists from relevant sources - both those that are registered and those that are not, and whether more information can be obtained about the facilities and tenures offered in registered boarding houses.
Supported accommodation for the homeless
As outlined in above, the ABS obtains lists of homeless supported accommodation and uses a 'green sticker strategy'.
However, the nature of the provision of accommodation to homeless people in need of housing support services is changing. In the lead up to the next Census, procedures will be reviewed to ensure that people in new forms of supported accommodation are correctly classified. Advice from service providers will be used about clients who are accommodated using vouchers or 'brokerage' in hotels, caravan parks etc to classify them as people in supported accommodation for the homeless.
Transitional Housing Management Units
Long-term supported accommodation, often described as Transitional Housing Management (THM) Units, often provides some security of tenure, and the dwellings are adequate (including basic kitchen facilities and a bathroom) and the household has privacy and exclusive use of those basic facilities. The THMs that meet these levels of housing would, under the ABS statistical definition of homelessness, not be classified as homeless. However persons living in those THMs that lack one or more of these elements would be classified as homeless.
Unfortunately, the lists of supported accommodation that the ABS obtained for the 2006 Census did not separately identify THMs from other crisis accommodation, nor report information about the elements of homelessness. Therefore all THMs are indistinguishably included in the broader estimate of 'Persons in supported accommodation for the homeless'. For 2011, the ABS did receive information from some jurisdictions about THM but not obtain any of the requested information about the facilities or the tenures offered. The ABS is analysing the list information received for 2011 to consider ways in which it can better separately identify THMs from and other supported accommodation in future Censuses.
Census collectors are required to visit 'private' dwellings 5 times (at different times and on different days) to try to make contact.
Where contact cannot be made, collectors are encouraged to speak to neighbours, or a 'credible source' about whether the dwelling was occupied on Census night.
If the collector' is not 'absolutely certain' that the 'private' dwelling was not occupied on Census night, system imputation follows.
No imputation was undertaken for or is required for NPDs (e.g., boarding houses, or supported as determined by area supervisors to be NPDs and their classification described by the owner/manager).
No imputation is undertaken for 'rough sleepers', but is undertaken for improvised dwellings.
For the 2006 Census imputation, 766,000 persons were imputed to be in non-contact private dwellings.
For about a third of the imputed people (about 275,000), people were imputed where the collector obtained information that there were some people in the dwelling on Census night - the number and sex of the people is obtained. For these records, the collector record book information on dwelling type, structure and location, is used together with the third party reported numbers of people by sex, to find a 'donor' dwelling near by to replicate the age, sex, usual residence (or 'place of enumeration') and registered marital status of that donor dwelling onto the imputed household in the non-contact dwelling. No other variables are imputed.
The remaining two thirds (about 500,000 people) where no information could be obtained as to whether or not the dwelling was occupied on Census night, are fully imputed based on looking for a donor record based on dwelling type, structure and location.
Imputed records are retained in the estimates of homelessness for the group 'Persons in supported accommodation for the homeless' (SAAP). While some SAAP properties were enumerated as NPDs in the 2006 Census, most were enumerated as private dwellings and reassigned a SAAP flag based on the lists of SAAP properties provided by jurisdictions. In the flagged properties where no contact was made, about 860 people were imputed using a credible source to confirm the occupation of the dwelling on Census night. Another 1,110 people were imputed without a credible source. While only a third of the national rates of imputation were undertaken with a credible source, for SAAP the rate was closer to 50%.
For listed SAAP properties it is known that the property is not a second home, nor a holiday home or a vacant property up for either sale, demolition or renting etc. And for nearly half of the imputations a credible source is used and no more information is needed to confirm the homeless state of the people in the dwelling on Census night. For the 1,110 or so imputed from a 'donor' record, this estimate may understate or overstate the numbers of homeless people actually in those SAAP properties on Census night. However, the scale of any such error is not likely to be large nor biased (an estimation error on only about 6% of total SAAP accommodation on Census night) and the inclusion of the imputed records lifts the Census estimates slightly to align broadly with Australian Institute of Health and Welfare (AIHW) estimates of SAAP accommodation on Census night, suggesting no significant overestimate.
Imputed records for people staying in non-contact private dwellings are excluded from the homeless estimates because there is no certainty about the nature of the dwelling occupancy on Census night (it might be a second home, a holiday home or a vacant property up for either sale, demolition or renting etc.) and no information is available about the characteristics of the occupants to assess their likely homelessness.
Further detail on the treatment of imputed records for relevant groups is in Chapter 5.
Income is used to assist in classifying people as likely to be homeless. Across the different group, ABS has used individual income and household income, depending on the living situation being assessed. The income thresholds are adjusted for each Census to remain relevant with increases in income over time. More detail on the reasons for the use of particular income thresholds is included in the relevant groups in Appendix 1.
People not classified as homeless
ABS homelessness estimates relate to usual residents of Australia and do not include overseas visitors. Nor do they include people who were enumerated in offshore oil rigs, drilling platforms etc, or who were on board a ship in Australian waters or on an overnight journey by train or bus.
People enumerated in Other territories are not included in the homelessness estimates.
Key differences between methodology employed for estimating homelessness for different Census years
Income, mortgage and rent cut offs
The use of Income, mortgage and rent cut offs in the rules for estimating homelessness are adjusted for each Census year using data from Survey of Income and Housing and the Consumer Price Index where appropriate. The cut offs for 2006 and 2001 are outlined in the rules in Chapter 5. The cut offs for estimates for the 2011 Census will be published with the 2011 estimates in Census of Population and Housing: Estimating Homelessness, 2011 (cat. no. 2049.0). See Appendix 1 for more information on the cut offs for income, mortgage and rent for 2006 and 2001.
Persons in supported accommodation for the homeless
While the list and green sticker strategies for SAAP properties were undertaken in the 2001 the information was not retained. Therefore, for 2001 homelessness estimates of persons in supported accommodation for the homeless, the ABS will use data from the AIHW Supported Accommodation Assistance Program (SAAP) collection on the number of clients and accompanying children accommodated on Census night for every state and territory except Victoria. The Department of Human Services (Victoria) provided the ABS with a comprehensive list of their SAAP and THM properties, this was used along with the national SAAP data collection (to establish the number of women in domestic violence services) to provide an overall estimate of number of people in SAAP accommodation for Victoria in 2001.
However, the homelessness SAAP estimates for 2001 are only for totals and by State and Territory and no Census demographic characteristics are available. Nor is it possible to remove overlaps between the AIHW based estimate and the other homeless operational groups which can be directly estimated using Census data, which will result in some minor over estimation.
Persons in Boarding Houses
For homeless estimates of persons in boarding houses, in addition to using information about non-private dwellings classified as boarding houses in the Census, rules are applied to:
The same basic rules for 2001 and 2006 are consistently applied as a starting point. However for 2006, additional information was available from this Census about need for assistance with core activities and volunteering to further refine the rules used reclassify as boarding houses dwellings that were enumerated as private dwellings. The refinements aim to better classify dwellings that were more likely to be convents or monasteries (and other similar religious institutions) and to remove dwellings more likely to be facilities for the aged. These two data items can not be used for the 2001 estimates as they were not collected in the 2001 Census. And as noted above, any dwellings incorrectly classified as boarding houses in 2001 that were SAAP properties cannot be removed from the boarding house estimates. The 2001 estimates may therefore overstate the homeless in boarding houses.