2050.0.55.001 - Discussion Paper: Methodological Review of Counting the Homeless, 2006  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 31/03/2011  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All

Discussion Forums

The ABS is holding discussion forums about this review in each state and territory capital city throughout April and May 2011. View Discussion Forums.



An early finding from the review was the need to clearly distinguish between:

  • enumeration in the Census of people who may be experiencing homelessness; and
  • subsequent analysis of Census records to estimate the number of people enumerated in the Census who are likely to be experiencing homelessness.

Improving the quality of the enumeration of people who may be experiencing homelessness will improve the quality of the estimates of the homeless. The ability of subsequent analysis of Census data to produce valid estimates of homelessness also depends on the validity of assumptions made and a detailed understanding of the Census characteristics.

A good quality Census enumeration provides the potential for better quality homeless estimates. Some improvements are being made for the 2011 enumeration, and stakeholder feedback has provided several suggestions to further improve enumeration in future Censuses.


Much of the discussion during the review concerned populations of homeless people that are likely to either avoid being enumerated in the Census, or whose circumstances are such that the Census is likely to miss them during enumeration.

This review concluded that the analysis of Census data should concentrate only on those people who were enumerated. Transparent and separate estimation/imputation for any underenumeration of designated homeless groups should not be replaced by misclassification of the enumerated population, as the size location and characteristics of the misclassified Census records were likely to be very different to the underenumerated populations of interest.


The review found that the changes made in each cycle of Counting the Homeless (CTH) introduced discontinuities in the published results.

A consistent methodology has been proposed in this discussion paper.


The review found an apparent lack of transparency with some aspects of the published CTH results. First, some of the published methodology was not followed in the compilation of the CTH results. Second, the assumptions published with, and underlying, the CTH use of Census variables did not always reflect the intent or design of those Census variables. Third, assumptions about the interpretation of some of the Census data did not appear to stand up to closer scrutiny, but users were not aware of the potential for a lack of coherence between the CTH method and the underlying Census data.

This review discussion paper provides a transparent articulation of a method of estimating the number of people enumerated in the Census that are likely to be homeless.


The discrepancy between assumption and method described in the commentary on transparency also leads to issues of coherence. Some assumptions about a population may be in error but are either subject to very small error or very consistent error, and hence might be tolerated. However, this review found that some of the populations that were being misclassified as homeless in CTH were growing rapidly and were affected by business cycle and demographic trends that undermined the utility of the results for most purposes.


A focus in much of the discussion during the review has been on the number of homeless people serviced during a year – this appears to be a perspective on homelessness that some participants in the review think should be apparent in a Census night count of all people homeless on that night. This confusion in measurement was also discovered in one of the CTH components – people staying in supported accommodation (SAAP) – where CTH documented a Census night count in SAAP / but used a count of support periods that spanned Census night and which included accommodation at any time during the support period. This significantly overstated the count of homeless people in this category.

ABS sought expert input on information that related the annual service flow counts with a single night estimate. Based on data published by AIHW on SAAP services, the number of homeless people accommodated in SAAP over the year in 2006–07 was about six times the number accommodated on Census night, AIHW (2008). Similar ratios appeared to apply in Reconnect. If this six to one ratio is applied to the reviewed ABS counts of around 65,000, it implies that over a year, 340,000 to 440,000 people may experience at least one period of homelessness.

Results from the 2010 ABS General Social Survey (due September 2011) are expected to shed more light on the total numbers of people experiencing homelessness over a period of time.

No other independent source was revealed that would translate a Census night count into a likely count of people experiencing homelessness over a period of time, but ABS will again seek such input during consultation on the discussion paper findings.