TECHNICAL NOTE 1 MEASURING CHANGE IN POPULATION ESTIMATES OF ABORIGINAL AND TORRES STRAIT ISLANDER PEOPLES

POPULATION ESTIMATES

1 While each Census aims to count everyone in Australia on Census night, the size and scale of the data collection process make it inevitable that some people will be counted more than once, and some will not be counted at all. The combined effect of this is measured by the Post Enumeration Survey (PES), which historically always indicates that there has been a net undercount (that is, more people were missed than were counted more than once).

2 An adjustment for the net undercount and other factors is applied to the Census counts to produce population estimates. This adjustment process is particularly important for producing Aboriginal and Torres Strait Islander population estimates, because the number of Aboriginal and Torres Strait Islander people not counted in the Census is considerably higher than the number who are counted more than once, resulting in a relatively large net undercount.

3 The importance of changes in undercount is such that the analysis in this Technical Note specifically includes a population-estimates based analysis of the change in Census counts. The analysis mirrors that presented in Measuring Changes in Population between Censuses, but population estimates, not Census counts, are analysed. This provides a basis for also assessing the proportion of change that can reasonably be attributed to undercoverage and changes to PES methodology.

For the analysis of change in population estimates the following framework has been used:

Population estimate 1 + births registered between Census 1 and Census 2 - deaths registered between Census 1 and Census 2 +/- net migration between Census 1 and Census 2 +/- the impact of measurable changes in Post Enumeration Survey (PES) methodology +/- the impact of measurable changes in Census coverage +/- the impact of factors which cannot be explained by measurable components of population change or changes in methodology or coverage = Population estimate 2

MEASURABLE COMPONENTS OF CHANGE IN POPULATION ESTIMATES OF ABORIGINAL AND TORRES STRAIT ISLANDER PEOPLES

Demographic factors

4 Analysis of the change in Census counts is limited by the fact that, unlike population estimates, adjustments are not made for undercount and any other factors (such as people temporarily overseas on Census night). For more detailed information about these adjustments, refer to Australian Demographic Statistics (cat. no. 3101.0). For this reason, population estimates are considered to be a more accurate reflection of the population than Census counts, and provide a more complete picture of ‘explainable' change.

5 Change attributable to demographic factors is presented here using estimated resident population (ERP) figures. It is important to note the data used in this analysis is based on registered births and deaths, rather than projected births and deaths. For more information, see the chapter Measuring Changes in Population between Censuses.

6 Analysis of the ‘demographic’ components of change in the estimated resident population reveals that a smaller percentage of the change observed in the population estimate of Aboriginal and Torres Strait Islander people between 2006 and 2011 is attributable to births, deaths and migration than was reflected in the analysis of the change in Census counts. This highlights the contribution of changes in Census undercount and the methodology used to measure undercount between the 2006 and 2011 Censuses, which are explored further in this chapter. It also reflects a change in the size of the adjustment for residents temporarily overseas which, given the relatively small impact, is not separately identified within this analysis.

	Population estimate		Relative change					Relative change due to demographic factors
	2006	2011	2006-2011		Natural increase(a) 2006-2011	Overseas migration 2006-2011	Interstate migration 2006-2011	2006-2011
	no.	no.	no.	%	no.	no.	no.	no.	%
New South Wales	152 685	208 476	55 791	36.5	18 056	411	-1 996	16 471	29.5
Victoria	33 517	47 333	13 816	41.2	4 629	174	833	5 636	40.8
Queensland	144 885	188 954	44 069	30.4	20 360	435	1 212	22 007	49.9
South Australia	28 055	37 408	9 353	33.3	3 732	57	326	4 115	44.0
Western Australia	70 966	88 270	17 304	24.4	9 356	172	294	9 822	56.8
Tasmania	18 415	24 165	5 750	31.2	1 810	37	-51	1 796	31.2
Northern Territory	64 005	68 850	4 845	7.6	5 582	44	-793	4 833	99.8
Australian Capital Territory	4 282	6 160	1 878	43.9	618	45	150	813	43.3
Australia(b)	517 043	669 881	152 838	29.6	64 160	1 375	. .	65 535	42.9
(a) Registered births minus registered deaths. (b) Includes Other Territories. Note: Totals and components may not be consistent within and between tables due to introduced random error to protect confidentiality of Census respondents - see Census Dictionary, 2011 (cat. no. 2901.0).

Changes in Census undercount and the method used to calculate undercount

7 Applying estimates of net undercount from the PES to the Census counts is a major part of the calculation of the estimated resident population. While the above analysis of population change components provides important insight into the increase between the 2006 and 2011 Aboriginal and Torres Strait Islander population estimates, it is essential that any change analysis includes an estimate of the change in undercount over this period. It is also important to consider any measurable impact of a change in PES methodology, which will affect the comparability of undercount estimates over time.

8 Two key changes were made to the 2011 PES methodology which had an impact on the comparability of 2006 and 2011 net undercount figures for the Aboriginal and Torres Strait Islander population. The first of these changes was an improvement in the collection of Indigenous status in the PES and is likely to have had the greatest impact. A household-level question (used in 2006) was replaced with a person-level question in the 2011 PES. This meant that the Indigenous status question was collected for everyone in the dwelling on a person by person basis, resulting in identification in the 2011 PES that was more closely aligned with that in the Census. For more information on this change, refer to Technical Note 1: Improvement in Collection of Indigenous Status in Census of Population and Housing - Details of Undercount, 2011 (cat. no. 2940.0). Had this improved methodology been available in 2006, it is estimated that the net undercount of Aboriginal and Torres Strait Islander people in 2006 would have been much higher than the published 11.5%, and likely higher than the 2011 net undercount of 17.2%.

9 The second methodological change in the 2011 PES was the introduction of Automated Data Linking, which significantly improved the PES linking and matching methodology and, consequently, the net undercount estimates. If it had been possible to better match PES and Census records for Aboriginal and Torres Strait Islander people in 2006, this methodological change would have resulted in a small decrease in the net undercount in 2006, a small offset to the larger impact from the change in identification.

10 The challenge with estimating these impacts is the need to involve assumptions. Some assumptions have been required, given there is no possibility of repeating the collection of Indigenous status for the Census population in 2006, or of undertaking the linking and matching of 2006 PES and Census records since name and address information is destroyed at the conclusion of Census processing. This means that the ABS has produced its 'best estimate' of what the impact would have been in 2006, using the relationship in available data for the 2006 PES and 2011 PES. For this reason, these estimates should be considered illustrative of the impact of the change in PES methodology.

11 Almost one-quarter (23% or 35,600) of the 152,800 increase in the estimated Aboriginal and Torres Strait Islander population can be explained by changes in PES methodology, with a further 13% (19,500) reflecting an illustrative estimate of the change in the 'actual' Census undercount between 2006 and 2011. The combined effect of these changes was a net increase of 55,100 Aboriginal and Torres Strait Islander people between 2006 and 2011, which accounts for 36% of the increase in the population estimates between 2006 and 2011.

	Population estimate		Relative change		Changes to PES methodology(a)	Other changes in undercount(b)		Relative change due to methodology and undercount
	2006	2011	2006-2011				2006-2011
	no.	no.	no.	%	no.	no.	no.	%
New South Wales	152 685	208 476	55 791	36.5	13 596	6 551	20 147	36.1
Victoria	33 517	47 333	13 816	41.2	3 810	1 805	5 615	40.6
Queensland	144 885	188 954	44 069	30.4	8 918	5 693	14 611	33.2
South Australia	28 055	37 408	9 353	33.3	3 198	1 068	4 266	45.6
Western Australia	70 966	88 270	17 304	24.4	3 036	2 744	5 780	33.4
Tasmania	18 415	24 165	5 750	31.2	2 123	638	2 761	48.0
Northern Territory	64 005	68 850	4 845	7.6	726	636	1 362	28.1
Australian Capital Territory	4 282	6 160	1 878	43.9	284	224	508	27.1
Australia(c)	517 043	669 881	152 838	29.6	35 608	19 451	55 059	36.0
(a) Estimated net impact of measurable changes to Post Enumeration Survey (PES) methodology (that is, improved collection of Indigenous status and the introduction of Automated Data Linking). (b) Estimated change in undercount excluding the estimated impact of changes to PES methodology. (c) Includes Other Territories. Note: Totals and components may not be consistent within and between tables due to introduced random error to protect confidentiality of Census respondents - see Census Dictionary, 2011 (cat. no. 2901.0).

ERROR OF CLOSURE - POPULATION ESTIMATES OF ABORIGINAL AND TORRES STRAIT ISLANDER PEOPLES

12 Three-quarters (120,600) of the 152,700 increase in the estimated Aboriginal and Torres Strait Islander population between 2006 and 2011 can be attributed to known and measurable sources of population change, as well as the estimated impact of changes to PES methodology and Census coverage. The residual portion of the increase (32,200 people) represents 4.8% of the 2011 estimated resident population, slightly lower than the 6.1% error of closure observed when looking at the change in Census counts.

	Population estimate		Relative change		Total explainable increase(a)		Error of closure
	2006	2011	2006-2011		2006-2011
	no.	no.	no.	%	no.	%	no.	%
New South Wales	152 685	208 476	55 791	36.5	36 618	65.6	19 173	9.2
Victoria	33 517	47 333	13 816	41.2	11 251	81.4	2 565	5.4
Queensland	144 885	188 954	44 069	30.4	36 618	83.1	7 451	3.9
South Australia	28 055	37 408	9 353	33.3	8 381	89.6	972	2.6
Western Australia	70 966	88 270	17 304	24.4	15 602	90.2	1 702	1.9
Tasmania	18 415	24 165	5 750	31.2	4 557	79.3	1 193	4.9
Northern Territory	64 005	68 850	4 845	7.6	6 195	127.9	-1 350	-2.0
Australian Capital Territory	4 282	6 160	1 878	43.9	1 321	70.3	557	9.0
Australia(b)	517 043	669 881	152 838	29.6	120 594	78.9	32 244	4.8
(a) Change explainable due to births minus deaths, changes to Census processing, overseas and interstate migration and changes to Post Enumeration Survey (PES) methodology. (b) Includes Other Territories. Note: Totals and components may not be consistent within and between tables due to introduced random error to protect confidentiality of Census respondents - see Census Dictionary, 2011 (cat. no. 2901.0).