Australian Population Association Conference, Alice Springs - W.D. Borrie Lecture, 30 June 2008

'Australian Demographic Statistics in the 21st Century - The Challenges'

By Brian Pink, The Australian Statistician, Australian Bureau of Statistics


I would like to acknowledge the Central Arrernte people as the traditional owners and custodians for Alice Springs.


2 It is a great honour and a pleasure to be invited as Australian Statistician to present the 2008 W. D. Borrie Lecture. I do not profess to be a demographer unlike some of my predecessors who were leaders in the field. In fact the first Commonwealth Statistician, Sir George Handley Knibbs, was the author of The Mathematical Theory of Population, an internationally acknowledged treatise on demographic techniques. And of course Knibbs successor, Charles Henry Wickens, was similarly well regarded as a 'vitals' statistician and was recognised as the prime driver behind the release of the Bureau's first life tables in the 1911 Census of the Commonwealth of Australia Statistician's Report released in 1917.

3 In preparing for this presentation, I reviewed the 2000 Borrie Lecture by Emeritus Professor Jerzy Zubrzycki "The Borrie Legacy: a Foundation for an Australian Population Policy" which reflected on the contribution Borrie made to Australian population policy. Two things struck me. First, the contribution of Borrie to demography was very significant as evidenced by his Population Trends and Policies and his work on the National Population Inquiry. The second was that Borrie had a keen understanding of the importance of good quality statistics and that those statistics, like a light pole, should be for the purpose of illumination rather than merely support.

4 Although not a demographer by training, I am acutely aware as Australian Statistician of the importance of demographic statistics. In its most mundane form, it is a simple fact that in 2008-09 the Australian Government will be distributing to the states and territories in excess of $55bn in the form of GST payments, Health Care Grants and Financial Assistance Grants to support local government, and the magnitude or share of these payments and grants are largely dependent on the population estimates for the states and territories. Specifically the GST payments are based on the population estimates at 31 December 2007 as determined by me on 5 June 2008. These are but illustrative but they do highlight that at all levels of government, very large pots of funding are allocated on the basis of population estimates.

5 At the other end of the spectrum, it is demographic statistics that are at the heart of the Treasury's Intergenerational Report. This report focusses on the implications of demographic change for economic growth and assesses the financial implications of continuing current policies and trends over the next four decades, which in some respects is not dissimilar to Borrie's Report of the National Population Inquiry.

6 More generally, demographic statistics inform on the population of today, for example we are as a population getter older, we are living longer, we are having fewer children than generations past, although that appears to be turning around at present, women are having children later and there are much larger migration flows. These in turn inform on the population of tomorrow. In brief, demographic statistics are critical to informed decision making by government, business and community.

7 Today's lecture is titled "Australian Demographic Statistics in the 21st Century - The Challenges". While I would be the first to admit that there are, and continue to be, many challenges for demographic statistics, today I want to focus on two key contemporary statistical measurement challenges. The first, relates to the demography of Aboriginal and Torres Strait Islander Australians, and the second, to demography at the local level i.e. small area demography.

8 Both are areas of demographic statistics where the information needs have been, and continue to be, extensive and growing. At the same time the information supply has been of variable quality due to the significant measurement issues associated with each of them and availability of information from the various data sources has also been an ongoing issue. Finding solutions to these challenges is essential since the reality is that the demand for high quality information in these areas is increasing and I can predict with confidence will only continue to increase as we move through the 21st century.

9 It is against this background that the balance of this presentation today will provide some context, discuss these challenges and in turn explore some of the possible solutions to those challenges. As both these issues are also discussed in more detail in separate papers being presented at this Conference, my focus will be on the strategic issues relating to these topics and where the ABS is going in addressing these issues.

10 In exploring those solutions, I want to emphasise that the ABS clearly understands that it is not the only, or in various cases the major source of relevant information. In fact we have, as an organisation, been strongly promoting the concept of a national statistical service recognising that there are many organisations that have information that can inform decision making and it is in that context I am strongly encouraging and supporting other agencies, both at the Commonwealth and state/territory level, to compile and expose their data. For example, Indigenous Australians have probably the highest compliance burden, in terms of participating in surveys and research of any sub population in Australia. They are surveyed more than anyone and hence there must be a rich information store in the various administrative bodies around Australia, and I have no doubt that some of these administrative datasets have the potential to greatly inform on the demography of Aboriginal and Torres Strait Islander Australians.

Some Context - The Current State of Play

11 However, before discussing those specific challenges I thought it useful to provide an overview of the current state of play of demographic statistics in Australia.

12 Since its formation in 1905 the ABS (or the Commonwealth Bureau of Census and Statistics as it was then known) has had a strong focus on demographic statistics. In part that reflected the importance of population estimates in the context of the colonies and a new nation, but it also reflected the strong influence of the above mentioned Knibbs and Wickens as the first two Commonwealth Statisticians.

13 As a result, the coverage and quality of ABS demographic statistics today is of a high standard and internationally recognised. It is a position we have, as an organisation, worked hard to achieve and it is one we want to maintain. It is not a position enjoyed by all statistical agencies. For example, our British counterparts at the Office of National Statistics have recently been the subject of some strong criticism in the UK House of Commons Treasury Committee report 'Counting the Population'. Our colleagues at the US Bureau of the Census are also under considerable ongoing criticism as they prepare for their next decennial Population Census in 2010.

14 Apart from our strong history in terms of demographic statistics there have been a number of contributing factors to our current position:
  • regular refreshing of population benchmarks through the conduct of the population census on a five yearly basis,
  • good quality births/deaths data sourced from the Registrars in each of the states and territories,
  • the island state nature of Australia and the excellent records of movements of persons into and out of Australia,
  • access to a range of other administrative data eg Medicare data and Australian Electoral Commission data, and
  • good ongoing relationships with the bodies responsible for these administrative data sources.

15 The two areas of particular strength for Australia in the international context are the 5 yearly re-benchmarking of population estimates to the census - many countries conduct their population census with a 10 yearly frequency; and the excellent data sources in respect of international migration and all movements across our borders which is becoming an increasingly significant but difficult area in many countries and was the source of much of the criticism in the UK House of Commons Treasury Committee report 'Counting the Population' mentioned earlier.

16 If asked to describe the current state of play in respect of demographic statistics in Australia I would describe it as excellent at the national level, and very good at the state/territory level. However, in relation to the specific themes I am discussing this evening, I would suggest that demography at the local level is generally of acceptable quality although I accept there is considerable scope for improvement in that quality. In respect of the demography of the Aboriginal and Torres Strait Islander Australians, my assessment would be that there are significant measurement, coherence, comparability and consistency issues, and it is those issues that we are trying to address.

The Demography of the Aboriginal and Torres Strait Islander Australians

17 Policies in respect of Aboriginal and Torres Strait Islander Australians have been, and continue to be, a high priority at all levels of government. The new federal government has strongly re-affirmed that high priority, and specifically COAG has identified a number of targets related to Indigenous well-being. These are to:
  • Close the life expectancy gap within a generation
  • Halve the gap in mortality rates for Indigenous children under five within a decade
  • Halve the gap for Indigenous students in reading, writing and numeracy within a decade
  • Halve the gap in employment outcomes between Indigenous and non-Indigenous Australians within a decade
  • All four year olds in remote communities to have access to early childhood education within five years
  • At least halve the gap for Indigenous students in Year 12 attainment or equivalent attainment rates by 2020

18 The Prime Minister has announced that he will report annually to Parliament on progress against the first three.

19 The need for good quality information on the demography of Aboriginal and Torres Strait Islander Australians is important to inform on the first two aspects. More generally the ABS recognises the need for good quality information to assist the decision making process at all levels of government and by the community.

20 In that regard, as noted earlier, there are a number of significant measurement issues in respect of the demographic statistics of Aboriginal and Torres Strait Islander Australians, namely:
  • the quality of the 5 yearly benchmarks from the census, which are compromised by the variable coverage of Indigenous Australians in the census, and the capacity of the Post Enumeration Survey (conducted directly after the Census) to adequately measure the undercount and the variability of the coverage, with an adequate degree of accuracy;
  • the changing propensity of people to identify as Aboriginal and/or Torres Strait Islander Australians which can make reconciliation between the respective census benchmarks problematic;
  • the quality and variability of Indigenous identification on birth and death registrations. In particular the latter, where the available evidence suggests that around only 55% of Indigenous deaths are being identified and recorded as such, in combination with changing propensity to identify, make the compilation of life tables and in turn life expectancy estimates an almost intractable problem.

21 Compounding the above issues, albeit to a lesser degree of significance, is the need to develop measures of overseas migration, interstate migration and intrastate movements so as to improve the quality of state/territory and lower level estimates of Indigenous populations for post censal years.

22 As a result of these measurement issues it must be said that the demographic statistics of Aboriginal and Torres Strait Islander Australians have quality issues and these are amplified as you drill down to the finer levels ie by geography or age. Furthermore they make the construction of life tables, estimates of life expectancy and population projections with any degree of accuracy, a significant challenge.

23 Ultimately these limitations have the opportunity to manifest themselves in terms of poor decisions as it relates to service delivery and/or policy development.

24 While the identification of the measurement issues are relatively straightforward, the same cannot be said about the resolution of said issues, and unfortunately there are no quick fixes. Indeed if there was we would have fixed them! In terms of the way ahead for ABS we will be focussing on the following in the short to medium term:
  • the key to good quality demographic statistics of Aboriginal and Torres Strait Islander Australians is getting the 5 yearly benchmark right. In that regard the 2006 Census was a step in the right direction in relation to enumerating the Indigenous population
  • in terms of getting the 2011 Census right, the prime objective is to get high quality and consistent coverage of the target population. To that end we will:

- extend and expand our Indigenous Engagement Strategy,

- more generally we will be developing a Northern Australia 2011 Census collection strategy which will include working more closely with local government authorities and organisations in hard to enumerate areas,

- a particular aspect of our strategies will be a strong focus on working more directly and closely with Aboriginal and Torres Strait Islander communities and organisations,

- further build on the close engagement we had with state/territory governments in 2006
  • complementing the census is the Post Enumeration Survey. The 2006 Post Enumeration Survey strategy was a major development in that its scope was expanded to include remote areas and discrete Indigenous communities. The results from the PES portrayed a picture of variable and at times significant undercount of Aboriginal and Torres Strait Islander Australians in the Census, with a net undercount rate of 11.5%. Having said that, that picture is of itself of limited quality
  • we have commenced planning for the 2011 Post Enumeration Survey, which will build on the 2006 PES experiences
  • continue to work with Registrars in the states and territories, and in turn their data suppliers, to improve the completeness and timeliness of Indigenous birth registrations but more so to improve the Indigenous identification on death registrations. The latter presents a significant stumbling block to compiling quality population estimates in the intercensal years. The absence of good deaths and population count data presents major difficulties in compiling life expectancies and population projections and may compromise the ability of the Prime Minister to report with confidence to Parliament on how the nation is progressing in closing the gap on Indigenous disadvantage
  • we also need to do work to better understand the reasons for different coverage levels in deaths data in different jurisdictions so that future efforts can be better targeted. The ABS Census Data Enhancement Quality Study on Indigenous Mortality will be informative - we expect results to be released in November 2008. There are also increasing opportunities to use linked administrative data to understand whether or not the same people are being identified as Indigenous in different administrative collections
  • we will work with other government agencies to identify additional sources of data and strategies within jurisdictions to improve identification where it is an issue
  • we will continue to engage with the experts and users for advice - that's why I and a number of ABS officers are here, and it is the reason that we are active participants in the annual Australia and New Zealand Population Workshop and more generally the APA!

Demography at the Local Level

25 The second matter which I want to flag as a particular challenge for the 21st century is that of demography at the local level or small area demography. By demography at the local level we are referring to sub-state data down to the local government area or statistical local area or indeed lower if users were to have their way. Good quality information at this level is particularly relevant to regional policy considerations by all levels of government, and funding decisions and service delivery at the local level in all its forms ie government, business and community. The information needs become even more complex when there is a demand, as is often the case, for Indigenous/non-Indigenous and other sub populations, such as age/sex and non-English speaking, breakdowns of population estimates at the local level.

26 As for demographic statistics of Aboriginal and Torres Straits Islander Australians, demography at the local level is subject to a range of measurement issues. There are two elements to the measurement issue. The first is that of getting the benchmarks provided by the census right; the second is measuring changes in the intercensal period and I will address each of these in turn.

27 In relation to the question of census benchmarks, the issues are the extent of undercount in the census, the capacity of the Post Enumeration Survey to adequately measure that undercount and the appropriateness of applying these broad measures of undercount at fine geographical level. In simple terms undercount in the census is not an insoluble problem; rather it is the variability in the undercount across sub groups in the population and geography, and the capacity of the Post Enumeration Survey to capture that variability, that presents the issues for demography at the local level. And the reality is that there is variability in the undercount. While overall the net undercount in the 2006 census of 2.7% was higher than past censuses, it is an excellent result by world standards. However the undercount varies by state and territory eg Northern Territory was 7.6%, Queensland 3.7% and the ACT 1.2%; it varies by sex with males being 3.3% and it varies by age with the 20-24 and 25-29 year olds being hard to capture with an undercount of 6.8% and 6.9% respectively. In addition to these complexities we have to allocate over 350,000 residents temporarily overseas on census night to their respective local level geography.

28 While getting the benchmarks right is a difficult issue, it pales somewhat when compared to the issue of intercensal measures as they apply to demography at the local level. There are a range of factors that influence our capacity to compile accurate measures of demography at the local level in the period between censuses. These include:
  • settlement/departures of immigrants/emigrants. Net overseas migration accounted for 56% of Australian population growth in 2007 and while we have accurate measures of the nearly 600,000 'migration' movements (ie both immigrants and emigrants) resulting in net overseas migration of some 180,000 persons at the national level there is only limited information of where immigrants settle and where emigrants leave from. Our expectations are that the volume of net overseas migration will only increase as demand for labour in the global market increases;
  • measures of interstate migration. While the measures of net interstate migration which are largely based on Medicare change of address information are adequate at the broad level ie state and territory, the reality is that there are over 350,000 people moving interstate every year;
  • measures of intrastate movements ie from one local region to another within the same state or territory - in the 2006 census 2.8 million Australians reported that they lived at a different address one year ago.

29 The above measurement issues mean that it is necessary to model estimates below the state/territory level. Essentially that is achieved by compiling estimates at the state and territory level and assigning population growth within each of the states and territories to lower levels using other indicators of population growth, specifically Medicare data, Australian Electoral Commission data and building approvals, with the constraint that the component lower level estimates must equal the state or territory estimate.

30 In spite of the difficulties identified above, our analysis of estimates at the statistical local area based on the 2001 census benchmarks and those using the 2006 census benchmarks show that in the main they are not too dissimilar. It is the subject of another paper at this conference but for SLAs the average absolute difference was 3.75%, a decrease on the 2001 average of 3.82%. Of course that is little consolation when your population has been underestimated for a five year period and that argument is equally applicable at the state or territory level as it is to demography at the local level.

31 Those measurement issues are exacerbated when users want estimates at very low levels such as collector's district and/or non ASGC geographies (eg. post code, electorates, catchment areas) or want an Indigenous/non-Indigenous breakdown of the estimates. A further dimension to the problem is the need for information about service populations as opposed to usually resident population.

32 While the information needs are valid, particularly from a service delivery perspective the capacity to respond is limited. Specific initiatives the ABS will be pursuing in the short to medium term are:

(a) as for demographic statistics of Aboriginal and Torres Strait Islander Australians, the key to getting good quality demographic statistics at the local level is getting the 5 yearly benchmark right. As noted that involves two steps. Firstly, minimising the undercount and second accurately measuring the extent and variability of the undercount. In respect of the first of these we will be reviewing our Census awareness and promotional campaigns and collection strategies so as to maximise the response to the census. In relation to the Post Enumeration Survey the 2006 collection was a major improvement in terms of measuring the undercount and as noted earlier we will be building on that strategy for 2011

(b) working with agencies at all levels of government to identify additional administrative data sources so as to better measure net internal migration and overseas migration at the sub-state level. For example we are currently assessing the possible use of additional data from the Department of Immigration and Citizenship to further inform on areas of population growth and decline

(c) subject to b) we will be looking to review our modelling processes for allocating growth within states and territories. This will include assessing whether we have the necessary information and capacity to compile broad sub state (eg capital city/rest of state) estimates and then model down to the finer levels from that level, rather than from the state or territory level

(d) extending the use of administrative data sources into analyses of internal migration flows

(e) ABS recently released an Information Paper on Population Concepts to provide a framework of population concepts including the service population. The next step will be looking at the information needs for data using these different concepts and the capacity to address those needs.


33 At the outset of this presentation I noted that there are and will continue to be many challenges for demographic statistics in the 21st century. The two topics I have addressed tonight ie the demographics of Aboriginal and Torres Strait Islander Australians and Demography at the Local Level are two challenges where the ABS recognises that the quality is variable and the need for high quality data to enable informed decision making by government, business and the community is increasing.

34 The objective of my presentation today has not been to present excuses and consign these complex issues to the too hard basket. Rather it has been:
  • to ensure there is a common understanding of the issues, and how they might impact on the data available ie. its fitness for purpose;
  • to outline the steps the ABS is taking to address the measurement issues we are confronting;
  • to enlist practical and tangible support to ensure a successful 2011 Census, which is the lynchpin of our demography statistics; and
  • to encourage others to bring to attention and open the lid on other data sources and strategies that may be available and which might assist in addressing the data gaps.

Thank you.