3126.0 - Demography Working Paper 2001/4 - Issues in Estimating the Indigenous Population, 2001
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 27/08/2001
Page tools:
 Print Page
 Print All
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
OUTLINE 5. ISSUES
5.2 The propensity to identify as Indigenous being different over time and between collections 5.3 The existence of people for whom their Indigenous status is not known 5.4 The problems of assessing mortality given incomplete data 5.5 The measurement of net census undercount 6. CONCLUSION 7. REFERENCES For further information, contact Shahidullah at m.shahidullah@abs.gov.au 1. INTRODUCTION The ABS has been involved in producing experimental estimates of the Australian Aboriginal and Torres Strait Islander (Indigenous) population since 1994 when it published Experimental Estimates of the Aboriginal and Torres Strait Islander Population, 1986-1991 (ABS, 1994; ABS, 1998a). It included estimates for 1986. Subsequently a 1991-1996 issue of this publication was released. Estimates were published at the State (including Territory) level and, for June 1996, at the Aboriginal and Torres Strait Islander Commission (ATSIC) region level. Estimates of the Torres Strait Islander population at June 1996 have been published in Australian Demographic Statistics, June Quarter 2000 (ABS, 2000b). Experimental projections of the Indigenous population have also been published at the State level in the 1991-2001 and 1996-2006 issues of Experimental Projections of the Aboriginal and Torres Strait Islander Population (ABS, 1996; ABS, 1998b). Projections, which involve a series of assumptions about future trends in fertility, mortality, migration and the propensity to identify Indigenous origin on census forms, provide the best indication of post-censal populations in the absence of post-censal population estimates. Estimates of the Indigenous population are 'experimental' in that the estimates are affected by changes in the propensity of people to identify as being of Indigenous origin on census forms and satisfactory data on births, deaths and migration are not generally available. 'Experimental' reflects the uncertainty involved. The principal focus of this paper is on estimates for 1991-1996 based on the 1996 Census. 2. WHO IS AN INDIGENOUS PERSON? A definition of an Aboriginal or Torres Strait Islander person was given in a High Court judgement in the case of Commonwealth v Tasmania (1983) 46 ALR 625. This definition states that an Aboriginal or Torres Strait Islander is a person of Aboriginal or Torres Strait Islander descent who identifies as an Aboriginal or Torres Strait Islander and is accepted as such by the community in which he or she lives. The population census question asks 'Is the person of Aboriginal or Torres Strait Islander Origin?'. This question measures the descent concept although some respondents will interpret the question to mean both descent and identification. It does not take account of the third part of the definition, community acceptance. 3. SIZE, DISTRIBUTION AND CHANGE OF THE INDIGENOUS POPULATION The size and distribution of the Indigenous population reflected in experimental estimates at the State and Territory level are shown in the following table. 1. Estimated Resident Indigenous Population, 30 June 1986, 1991 and 1996
Source: ABS: Experimental Estimates of Aboriginal and Torres Strait Islander Population, 1986-1991 and 1991-1996 (ABS, 1994 and ABS, 1998a). The Indigenous population has increased substantially over the ten years from 1986 to 1996. The percentage increase of the Indigenous population from 1986 to 1991 and from 1991 to 1996 are shown below in table 2. 2: Percentage Increase in Estimated Resident Indigenous Population, 1986-91 and 1991-96
(b) 1991 estimate based on the 1991 Census. 1996 estimate based on the 1996 Census. Source: ABS: Experimental Estimates of the Aboriginal and Torres Strait Islander Population, 1986-1991 and 1991-1996 (ABS, 1994 and ABS, 1998a). The Indigenous Estimated Resident Population (ERP) increased from 283,000 in 1991 to 386,000 in 1996 (Table 1). This represents an increase of 36.7% (Table 2). The increase in ERP is higher than the 33% increase in Indigenous census counts between 1991 and 1996 (Ross, 1999). This difference is associated with the different methods used to derive ERP in 1991 and 1996 (see Section 4). 4. HOW THE INDIGENOUS POPULATION WAS ESTIMATED 4.1 PRODUCING ESTIMATES Experimental estimates of the Indigenous population, 1991-96, were produced in two stages. Firstly, various adjustments were applied to census counts to produce the estimated Indigenous population at 6 August 1996 (census date). Secondly, these 6 August estimates were 'survived' back to 30 June 1996 and then back to 30 June 1991 using life tables (the reverse survival method).4.1.1 Census date estimate The starting point for a Census date (6 August 1996) Indigenous estimate was usual residence counts for Indigenous people from the 1996 Census of Population and Housing . Adjustments were then made for:
(a) Persons living in the Australian external territories of Christmas Island or the Cocos (Keeling) Islands were excluded on the basis that they were highly unlikely to have been Indigenous (b) Country of birth of parents In the 1996 Census of Population and Housing, some people identified as being of Aboriginal or Torres Strait Islander origin who had both parents born overseas. It was considered that Indigenous people where both parents were overseas born were highly unlikely to have been Indigenous. A person was deemed to have been born overseas if their country of birth was neither 'Australia' nor 'Not Stated'. (c) Non-response to the census question There were two circumstances under which a 'not stated' response was recorded to the Aboriginal and Torres Strait Islander question on the 1996 Census form. The first was when a form was completed but there was no response to the Indigenous origin question. The second was when a census collector considered that a person did reside at a particular residence on census night, but was unable to obtain a completed census form for that person. This would happen because the either census collector didn't make contact, the respondent refused to complete a census form or the respondent told the collector they would mail back the form, but subsequently no form was received by ABS. In the second case, a 'dummy' census form was created containing imputed data for age, sex and state of usual residence on census night, but with all other characteristics set to 'not stated'. Table 1 shows the percentage distribution of 'dummy' category by Indigenous origin. The shaded area shows how the non-stateds are distributed. Of note is the fact that 55% of the not stateds resulted from a partially completed census form. This compares with 45% where no characteristics were obtained from a census form. 3: Dummy category by Indigenous satus at 6 August 1996 (%)
Source: ABS: 1996 Census of Population and Housing It was assumed that some of these 'not stated' responses were actually of Indigenous origin. An imputation procedure was applied to allocate Indigenous status to non-respondents according to the proportion of Indigenous persons in the total number of stated responses by:
(d) Net census undercount The 1996 Census Post-Enumeration Survey (PES) indicated that Indigenous people were four to five times more likely to be missed by the 1996 Census than non-Indigenous people (Table 2). 4: Net Undercount, 1996 Census Post-Enumeration Survey
(2) The standard error expressed as a percentage of the estimate. Source: ABS: Demographic Estimates and Projections: Concepts, Sources and Methods, table 4.5. The 1996 PES also indicated that net undercount rates varied significantly by age, sex and geographical area. For example, Indigenous males aged 20 to 29 years had one of the highest rates of undercount. However, because of the limited sample size of the PES (37,000 households in 1996), it was not possible to derive reliable estimates of Indigenous undercount for every cross-classification of age, sex and State/Territory. Instead, the age, sex and State/Territory undercount distribution for the total Australian population was applied to the total Indigenous undercount rate of 7.1 percent. The PES was not conducted in remote areas or Indigenous communities because special enumeration procedures were used in the census and the PES could not provide an independent measure of undercount. The national undercount was therefore applied to all geographic areas in the absence of other information. (e) Underreporting of infants The census counts at age zero for Indigenous persons were also compared with registered births during 1995-96. The number of live birth registrations in South Australia, Western Australia, Tasmania, the Northern Territory and the Australian Capital Territory was notably higher than the enumerated census population aged zero (Table 2). In these States/Territories the estimated population at age zero was increased to be more consistent with birth registrations. The population at ages 0, 1, 2, 3 and 4 was then smoothed for all States/Territories to ensure a satisfactory transition from age 0 to age 4, although the total population aged 0-4 years remained unchanged. 5: Adjustment to Indigenous Population Aged Zero
4.1.2 Effect of adjustments The effect of the above adjustments on the Indigenous population at 6 August 1996 for all states and Australia are summarised in Table 4. 6: Components of Estimating the Indigenous Population at 6 August 1996
(b) Includes Jervis Bay Territory (c) Includes adjustments to ages 0,1,2,3 and 4 based on demographic analysis of registered births data Source: ABS: Experimental Estimates of the Aboriginal and Torres Strait Islander Population, 30 June 1991 - 30 June 1996, (ABS, 1998a). 4.2 REVERSE SURVIVAL OF CENSUS ESTIMATES OF INDIGENOUS POPULATION After adjustments had been made to produce Indigenous population estimates at 6 August 1996, the estimates were backdated to 30 June 1991 (30 June being the standard reference date for population estimates). Usually the cohort component method would have been used, and this would have been based on demographic flow data (births, deaths and migration). However, satisfactory flow data for the Indigenous population were not generally available so Indigenous population estimates were backdated for each State/Territory by sex using the reverse survival method. This method involves backdating the adjusted census counts using the 1991-96 experimental Indigenous life table (see Appendix for more information). Zero net internal migration (by age, sex and State/Territory) and zero net overseas migration (by age and sex) was assumed. No attempt was made to adjust for any change in the propensity to identify as Indigenous.4.2.1 From 6 August 1996 to 30 June 1996 Firstly the estimated resident Indigenous population at 6 August 1996 was survived back one year to obtain an estimate at 6 August 1995. Next, an adjustment was applied to the estimated population aged 74 and 75+ years at 30 June 1996 to ensure a lower probability of survival for males compared with females. Finally the 30 June 1996 estimate was linearly interpolated between the 6 August 1995 and 6 August 1996 estimates, based on the number of days elapsed (37÷365). This resulted in a base population as at 30 June 1996.4.2.2 From 30 June 1996 to 30 June 1991 The estimated resident Indigenous population at 30 June 1996 was survived back one year at a time to obtain an estimate at 30 June 1991-95. To backcast the Indigenous population for ages 74 and 75+ using an Indigenous life table, the 75+ age group was first disaggregated into a single year age structure (75 to 99 years). This was done by applying the 75 to 99 age distribution of Indigenous census counts.4.2.3 Experimental Indigenous Life Table, 1991-96 The standard approach to calculating death rates relies on applying the number of deaths in a given period to the 'exposed to risk' population during that period. Death rates derived using this approach would under-estimate the actual level of Indigenous mortality as all Indigenous deaths are not registered as such. For this reason, ABS first estimated the completeness of Indigenous deaths registration using a method developed by Preston and Hill (1980). Then it inflated the Indigenous deaths registered during the 1991-96 intercensal period by the under-coverage factor derived from the Preston and Hill method to obtain an estimated number of Indigenous deaths. Finally, the estimated deaths were used to construct the experimental Indigenous life tables for the period 1991-96. The Preston-Hill method is based on the premise that in a 'closed' population (with no migration and no change in census coverage), the only cause of change in the size of a cohort is due to deaths. For a full exposition of the Preston-Hill method, see Preston and Hill (1980), and Luther and Rutherford (1988). For an update on more recent experimental Indigenous life tables see Deaths Australia, 1999 (ABS, 2000) and the ABS Demography Working Paper 2001/2 released in March 2001 (ABS, 2001). 4.3 1991 BASED ESTIMATES DIFFERENCE The method of estimating the 1991 Indigenous population was different from that of the 1996 population mainly in relation to the assessment of net undercount. To estimate 1991 Indigenous population from the 1991 Indigenous Census counts, the following adjustments were made (ABS, 1994):
A brief description of each adjustment follows: (a) ABS assumed that some of the people who did not respond to the Indigenous origin question in the census were actually of Indigenous origin. Some non-respondents were therefore allocated to being of Indigenous origin according to the ratio of the Indigenous to the total stated response population. This allocation was made separately for each sex at the Statistical Local Area (SLA) level. (b) The second adjustment was made to correct for errors in age reporting. Some age misstatement in the Indigenous population occurred in both the 1986 and 1991 Censuses. To remove the small distorting effects of age misstatement, a three term moving average was applied to the single year age distributions. This was done at the sex and State/Territory level. This adjustment resulted in a smoother age distribution for the Indigenous population, but did not affect the age by State/Territory totals. (c) The third adjustment was made to correct for undercount of the young adult Indigenous male population. Some evidence of underenumeration of young adult Indigenous males was found when sex ratios (the number of males per 100 females) between the 1986 and 1991 Censuses for age cohorts were compared. Sex ratios for some age groups in the Indigenous census counts were much lower than expected. ABS assumed that such low sex ratios was the result of male undercount rather than female overcount as female overcount in the census is unlikely. Sex ratio analysis was used to adjust for male undercount in the ages where the sex ratio was lower than expected. This involved adjusting the Indigenous male population at these ages to obtain the expected sex ratios for these ages. (d) The last adjustment involved correcting for net undercount of the Indigenous population. This adjustment was made by the female age-specific undercount rates for each State/Territory. These rates were derived from the 1991 Census PES for the total population. Reliable undercount estimates for Indigenous population were not available from this survey. 7. Components of Estimating the Indigenous Population at 6 August 1991
Source: ABS: Experimental Estimates of the Aboriginal and Torres Strait Islander Population, 30 June 1986 - 30 June 1991 (ABS, 1994). 5. ISSUES Issues associated with estimation of the Indigenous population involve the inconsistency of the official definition of an Indigenous person with what can be collected in the Census; the propensity to identify as Indigenous on forms being different over time and between collections; the existence of people for whom their Indigenous status is not known; the problems of assessing mortality given incomplete data and the measurement of net census undercount.5.1 Inconsistency of the official definition of an Indigenous person with what can be collected in the Census The population census question, responses to which provide the basis for population estimates, asks 'Is the person of Aboriginal or Torres Strait Islander Origin?'. This question measures the descent concept although some respondents will interpret the question to mean both descent and identification. It does not take account of the third part of the definition, community acceptance. Community acceptance can be a major factor. For example, the legal manager of the Tasmanian Aboriginal Centre Michael Mansell has said that 'there were more phoney than real Aborigines in Tasmania and more than half the voters in the 1996 ATSIC election were not Aboriginal' (Woolford, 2001). 5.2 The propensity to identify as Indigenous being different over time and between collections Because the concept of an Indigenous person includes social elements as well as a biological element, Indigenous people who have some non-Indigenous origins can be expected to vary in their identification, or propensity to identify, over time or in different statistical collections at the same time. Variations in identification can occur associated with perceptions of the service involved. By a person identifying as Indigenous, access to services may be perceived to be affected positively (eg rental assistance, access to public housing, health services, special education) or negatively (eg police and private rental). Census counts do vary over time. There was a 33% increase in census counts of Indigenous people between the 1991 and 1996 Censuses. There was wide variation between States and Territories ranging from an increase in census counts for Tasmania of 56% to 16% for the NT. Just over half of this increase can be explained by natural increase, changes in the transmission rate of Indigenous origin from parents to their children and changes in census editing procedures. While improvements in enumeration explain some part of the balance, most of the unexplained change is attributed to a change in the propensity of people to identify as Indigenous in the census (Ross, 1999). Despite this apparent large change in the propensity of people to identify as Indigenous in the 1996 Census, persons classified as Indigenous in the 1996 Census seem to have similar characteristics to Indigenous persons in the 1991 Census. (Hunter, 1998). Identification as Indigenous is inconsistent in different statistical collections run at much the same time. In the 1996 Census Post-enumeration Survey (PES) the number of people who were identified as being of Indigenous origin was a net 8% less than in the 1996 Census. The equivalent figure for the 1991 Census PES was a net 2% less and for the 1986 Census PES an additional 11% net (ABS, 1998). Changes in identification will not necessarily always result in an increase in identification. For example, New Zealand has experienced negative identification as Maori between 1981 and 1986 (Moore, 1989). 5.3 The existence of people for whom their Indigenous status is not known There were 353,000 persons counted in the 1996 Census as being Indigenous ( 2.0 % of the total usual resident count). This was greatly exceeded by the 525,000 persons for whom there was no answer recorded to the relevant question (3.0 % of the total usual resident count). The equivalent figures for the 1991 Census were 265,000 (1.6 %) Indigenous persons and 535,000 (3.2%) persons with not stated Indigenous status. While the non-response rate involved was below that for census questions on average, there is never-the-less a concern that the assumption involved in imputation procedure (to allocate Indigenous status to non-respondents according to the proportion of Indigenous persons in the total number of stated responses) may not be correct. 5.4 The problems of assessing mortality given incomplete data Accurate mortality assessment requires a good understanding of both the number of deaths and the population. In the case of the Indigenous population, this did not exist in the period 1991-1996. As a result, an Australia-wide mortality assessment was made for population estimate purposes using the Preston-Hill method to produce life tables. Two limitations with this approach exist. First, there is an implicit assumption that there is little variation between States in their mortality experience. Second, the Preston-Hill method assumes a closed population. While this is true in terms of Australian Indigenous people living in other countries (0.3% of Indigenous persons counted in the 1996 Census lived overseas 5 years ago), it is not correct in terms of the large number of persons who have been identified as an Indigenous person for the first time in the 1996 Census. ABS has commenced producing life tables at the State level and expects to have a far better base of knowledge to undertake 2001 Census based estimates. The State level life tables were published in Deaths, Australia 1999 (ABS, 2000). For an evaluation of this work see the ABS Demography Working paper 2001/2 (ABS, 2001). 5.5 The measurement of net census undercount The Census PES was used to provide the major part of a measure of the net census undercount. For Indigenous persons, the 1996 PES recorded a 7.1% net undercount with an associated relative standard error of 14.7%. There are several issues with this measure. First, the standard error, a measure of the chance of variation from the true figure, is relatively high. Second, the PES does not cover persons in discrete Indigenous communities and in sparsely settled areas. Discrete Indigenous communities, where special census procedures are used, were excluded from the PES because they often required the same enumeration contacts and procedures as used in the census, thus compromising the independent nature of the PES. For the NT this meant that only about 30% of its Indigenous population were subject to selection in the 1996 PES. On the other hand, NT's Indigenous census counts appear to be relativity good as reflected in the relatively small error of closure (Ross, 1999). There may also be a bias towards overcounting given the nature of the counting rule in such areas which unlike the rest of Australia includes persons not present at the census but who are judged will not be counted elsewhere. Third, the PES estimate of the Indigenous undercount is based on those who identify as Indigenous in the PES and not those who identify as Indigenous in the PES. Fourth, the PES is likely to miss the same people as the census misses. While the extent of this problem is considered to be very small for the total population given the close correspondence between PES results and estimates based on the 1991 Census updated for natural increase and migration, this may not be the case for the Indigenous population. 6. CONCLUSION There is a common theme associated with Indigenous population estimates and that is uncertainty. Because of the difficulties associated with estimating the Indigenous population, there is unlikely to be ever one answer to what the Indigenous population is in the absence of a definitive register of the Indigenous population. Effectively the population in technical terms is unbounded. For the analyst of Indigenous issues, it is important that there be an awareness of the assumptions involved in estimation. If an analyst does not take account of the uncertainty involved, for example not acknowledging the variation in population that could exist, then the conclusion from that analysis could well be flawed. 7. REFERENCES ABS, 1994, Experimental Estimates of the Aboriginal and Torres Strait Islander Population, 30 June 1986 - 30 June 1991, ABS Cat. No. 3230.0, Canberra. ABS, 1996, Experimental Projections of the Aboriginal and Torres Strait Islander Population, 30 June 1991 - 30 June 2001, ABS Cat. No. 3231.0, Canberra. ABS, 1998a, Experimental Estimates of the Aboriginal and Torres Strait Islander Population, 30 June 1991 - 30 June 1996, ABS Cat. No. 3230.0, Canberra. ABS, 1998b, Experimental Projections of the Aboriginal and Torres Strait Islander Population, 30 June 1996 - 30 June 2006, ABS Cat. No. 3231.0, Canberra. ABS, 2000a, Deaths, Australia 1999, ABS Cat. No. 3302.0, Canberra. ABS, 2000b, Australian Demographic Statistics, June Quarter 2000, ABS Cat. No. 3101.0, Canberra. ABS 2001, Aboriginal and Torres Strait Islander Mortality: Evaluation of Experimental Indigenous Life Tables, ABS Demography Working Paper 2001/2. ABS 2001, Demographic Estimates and Projections: Concepts, Sources and Methods. Hunter B, 1998 "Assessing the Validity of Intercensal Comparisons of Indigenous Australians", Journal of the Australian Population Association, Vol. 15, No. 1. Luther N Y and Rutherford R D, 1988, "Consistent correction of census and vital registration data", Mathematical Population Studies, Vol. 1, No. 1, pp. 1-20. Moore T, 1989, The 1981/86 Intercensal Consistency Study, Working Paper 1989/3, Mathematical Statistics Branch, Department of Statistics, New Zealand (Statistics New Zealand). Ross K, 1999, Occasional Paper, Population Issues, Indigenous Australians, 1996, ABS Cat. No. 4708.0, Canberra. Preston S and Hill K J, 1980, "Estimating the completeness of death registration", Population Studies, Vol. 34, No. 2, pp. 349-66. Woolford D, 2001, "Australia: Aboriginal Identity Problems", Townsville Bulletin, February 17, p.40. Document Selection These documents will be presented in a new window.
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
