Aboriginal and Torres Strait Islander life expectancy methodology

Propensities were calculated for death records that linked to Census records. Propensities were calculated for each combination of Indigenous status on the data sources as well as separately for the two age groups 0–59 years and 60 years and over. Due to low numbers of deaths in the 0–14 age group, deaths were aggregated for the age groups 0–14 and 15–59 to calculate propensities for the combined 0–59 age group. For death records where the Indigenous status was not stated, propensities were calculated for all ages combined.

Table 1 shows the numbers of linked deaths registered in the year after Census and Indigenous status in the Census. The propensities shown in Table 2 were calculated from the data in Table 1.

The following example refers to propensities calculated for death records with any Indigenous status in the age group 60 years and over in Table 2 that were linked to a Census response with Aboriginal and/or Torres Strait Islander status. For death records in the 60 years and over age group, the propensities were: 

• P(ATSI | ATSI) = propensity for a death record with Aboriginal and/or Torres Strait Islander status to be Aboriginal and/or Torres Strait Islander in the Census was 1,699/2,199 = 0.8018 
• P(ATSI | NI) = propensity for a death record with non-Indigenous status to be Aboriginal and/or Torres Strait Islander in the Census was 539/156,081 = 0.0035 
• P(ATSI | NS) = propensity for a death record with not stated status to be Aboriginal and/or Torres Strait Islander in the Census was 25/793 = 0.0315. This would be the same for 0–59 years, as only one propensity is calculated for P(ATSI | NS). 

To obtain the adjusted number of deaths by age group, each of the propensities in Table 2 were multiplied by the total number of deaths for the age group in the year after Census.

Adjusted deaths were calculated for the three age groups 0–14 years, 15–59 years and 60 years and over, as all three age groups were used in further calculations. Due to low numbers of deaths in the 0–14 age group, deaths were aggregated for the age groups 0–14 and 15–59 in Table 2. The propensity calculated in Table 2 for the 0–59 years age group was applied to the deaths with a particular death record Indigenous status, for both the 0–14 and 15–59 years age groups in Table 3.

The following example demonstrates the calculation to adjust deaths for the 60 years and over age group. The same was applied for the other age groups.

Refer to the propensities in Table 2, and total deaths in Table 3 for the 60 years and over age group:

• Adjusted deaths with Census Indigenous status of Aboriginal and/or Torres Strait Islander who were reported as Aboriginal and/or Torres Strait Islander in the death register = 2,815 x 0.8018 = 2,257
• Adjusted deaths with Census Indigenous status of Aboriginal and/or Torres Strait Islander who were reported as non-Indigenous in the death record = 177,063 x 0.0035 = 611
• Adjusted deaths with Census Indigenous status of Aboriginal and/or Torres Strait Islander who were reported as not stated in the death records = 785 x 0.0315 = 25.

The same calculation was applied for deriving adjusted deaths with Census Indigenous status of non-Indigenous and not stated using the propensities in Table 2.

The adjusted deaths for each Census Indigenous status, for each age group and death record status are shown in Table 4.

These adjusted deaths were used in the calculation of expected deaths (box 10) and age-specific adjustment factors (box 11).

Propensities for Census Indigenous status identification (box 4) were calculated for persons with Indigenous status of Aboriginal and/or Torres Strait Islander in PES, where PES data was linked to Census data.

Partial age-adjustment was applied at this step, combining age groups where the PES sample was too small to support accurate representation of that group.

The following example refers to data for the 60 years and over age group in Table 5.

• P(ATSI | ATSI) = propensity of PES Indigenous status being Aboriginal and/or Torres Strait Islander given Census Indigenous status was Aboriginal and/or Torres Strait Islander was 468/509 = 0.9194
• P(ATSI | NI) = propensity of PES Indigenous status being Aboriginal and/or Torres Strait Islander given Census Indigenous status was non-Indigenous was 269/102,654 = 0.0026
• P(ATSI | NS) = propensity of PES Indigenous status being Aboriginal and/or Torres Strait Islander given Census Indigenous status was not stated was 43/534 = 0.0805.

The above propensities are based on the unweighted PES data for simplicity in the example. The actual propensities were calculated using weighted PES data (box 4). The final propensities for the 60 years and over age group were:

• P(ATSI | ATSI) = 0.8967
• P(ATSI | NI) = 0.0024
• P(ATSI | NS) = 0.0673.

The expected number of deaths for each Census Indigenous status (box 10) was obtained by multiplying the propensities of identification for the age group (box 4) by the number of adjusted Census deaths in the same age group and Indigenous status (box 9). The expected deaths for each Census Indigenous status were then summed to get the total expected deaths. The following example refers to the calculation of expected deaths using data for the 60 years and over age group.

The adjusted deaths (box 9) and Aboriginal and Torres Strait Islander propensities for identification from Census-PES linked records (box 4) were used to calculate the expected number of deaths for persons aged 60 years and over:

= 2,893 * P(ATSI | ATSI) + 172,976 * P(ATSI | NI) + 4,793 * P(ATSI | NS)

= 2,893 * 0.8967 + 172,976 * 0.0024 + 4,793 * 0.0673

= 3,334.

This was the expected number of deaths to have occurred in the year after Census if the deaths were reported with the same Indigenous status as reported on PES.

See previous worked examples for input data. The expected number of Aboriginal and Torres Strait Islander deaths was 3,334 for the 60 years and over age group. The number of Aboriginal and Torres Strait Islander deaths in death records for the 60 years and over age group was 2,815. The identification rate for the 60 years and over age group was therefore:

= 2,815/3,334

= 0.844.

That is, it was estimated that the number of Aboriginal and Torres Strait Islander deaths in death records was underreported by a factor of 0.844. Only 84.4% of the Aboriginal and Torres Strait Islander deaths were being identified. To adjust the death records for this underreporting, the adjustment factor was taken as the reciprocal of the identification rate of Aboriginal and Torres Strait Islander deaths. The adjustment factor for the 60 years and over age group was:

= 1 / 0.844

= 1.184.

Death records are collected as administrative data from the Registrars of Births, Deaths and Marriages in each state/territory. All states and territories use information acquired from the Death registration form (DRF) to identify an Aboriginal and/or Torres Strait Islander death. While it is considered likely that most deaths of Aboriginal and Torres Strait Islander peoples are registered, a proportion of these deaths are not reported as such by the family, health worker or funeral director during the death registration process. Over time, jurisdictions have supplemented information from the medical certificate of cause of death (MCCD) to improve the identification of Aboriginal and/or Torres Strait Islander deaths in the data. Even with these improvements there remain a number of death records for which the Indigenous status is unknown; in 2022 this was 975 records (or 0.5% of all death records).

New South Wales started using information on the MCCD for identifying Indigenous status from 2022, and the deceased was identified as an Aboriginal and/or Torres Strait Islander person when recorded as such on either the DRF or the MCCD. If the two sources did not agree, identification on either source was given preference over recording the deceased as non-Indigenous. This change aligns deaths records for Aboriginal and Torres Strait Islander people in New South Wales with most other jurisdictions (with the exception of Victoria), where both the DRF and the MCCD have been used for a number of years. This change in how Aboriginal and Torres Strait Islander deaths were identified means that life expectancy estimates for 2020–2022 cannot be compared with previous life expectancy estimates for New South Wales.

The ABS has not previously released life expectancy estimates for Aboriginal and Torres Strait Islander people for Victoria and South Australia and they have not been released for 2020–2022. This document describes the reasons for not releasing these estimates.

A combination of quality indicators was used to determine which life expectancy estimates could be confidently released. This ensured that estimates were robust and fit for purpose.

Consistent with previous cycles, the following quality indicators were used for releasing the 2020–2022 life expectancy estimates:

• Average number of registered deaths of Aboriginal and Torres Strait Islander people by age group, sex and state/territory in 2020–2022
• Differences in classification of Indigenous status between Census and death records
• Aboriginal and Torres Strait Islander identification rates in death records
• Statistical error estimates.

To produce estimates of life expectancy for the Aboriginal and Torres Strait Islander population, deaths and population estimates were required for 19 age groups (0, 1–4, 5-year groups to 84, then 85 years and over) and sex. There was an average of 4,409 Aboriginal and Torres Strait Islander registered deaths in Australia during 2020–2022. Disaggregating these deaths by age groups and sex results in extremely small numbers of deaths for many age, sex and state/territory combinations.

For 2020–2022 deaths, of the 19 age groups:

• in New South Wales, one group had fewer than three average deaths for males and three for females
• In Victoria, four groups had fewer than three average deaths for males and six for females
• In South Australia five groups had fewer than three average deaths for males and eight groups had fewer than three average deaths for females.

Age-sex specific death rates and resulting life expectancy estimates based on small numbers of deaths in Victoria and South Australia would be unreliable.

The difference in classification was measured by the proportion of non-Indigenous death records that were identified as Aboriginal and/or Torres Strait Islander in Census, expressed as a percentage. It was calculated from Census and deaths linked data to assess the consistency of Indigenous status reported in Census and deaths records.

Differences in classification were:

• very low (6.4%) in the Northern Territory indicating highly consistent identification of Indigenous status in Census and death records.
• very high in Victoria (about 37%) and in South Australia (27%), indicating inconsistent identification in these states.  

The Aboriginal and Torres Strait Islander identification rates were calculated using Census and deaths linked data. They were measured by the ratio of the reported number of Aboriginal and Torres Strait Islander deaths to the expected number of Aboriginal and Torres Strait Islander deaths on a PES scale.

Producing identification rates involves calculating propensities of Aboriginal and/or Torres Strait Islander status in PES given Indigenous status in Census. For more information on how identification rates were calculated, see Calculating Aboriginal and Torres Strait Islander life expectancy.

Identification rates of less than one indicate an under-identification of Aboriginal and Torres Strait Islander deaths in death records relative to what would be expected on a PES scale. Identification rates of greater than one indicate an over-identification of Aboriginal and Torres Strait Islander deaths in death records relative to what would be expected on a PES scale.

In 2021-2022, identification of Indigenous status in death records was:

• high in Queensland (0.96) and in the Northern Territory (1.02)
• low in Victoria (0.54) and South Australia (0.84). 

The main considerations here were the PES sample size and associated sampling errors. The 95% confidence intervals for the PES propensities, identification rates and life expectancy estimates were all a function of these. Small PES sample leads to high PES sampling errors, which in turn leads to wide confidence intervals for propensities, death identification rates and life expectancy estimates.

The range of the confidence interval indicates how reliable an estimate is. A narrow confidence interval implies that the estimate is reliable, whereas a wide interval suggests the estimate is less reliable.

In 2021–2022, 95% confidence intervals were:

• relatively narrow for Australia and four jurisdictions that life expectancies were released for (that is, New South Wales, Queensland, Western Australia and the Northern Territory), indicating that propensities, identification rates and life expectancies were reliable for these geographies.
• very wide in the rest of the jurisdictions, including Victoria and South Australia, indicating estimates would be unreliable for these jurisdictions.

Victoria:

1. of the 38 age and sex combinations, 10 had less than three deaths in 2020–2022.
2. high difference in classification of Indigenous status in Census and deaths (37%)
3. low death Identification rate (58%)
4. 95% confidence intervals for propensity was much wider than those jurisdictions that life expectancies were released for (that is, New South Wales, Queensland, Western Australia and the Northern Territory). This indicated that propensity has high error bounds and high uncertainty. As errors in propensity flow through to identification rate and adjustment factor, resulting life expectancy estimates would also have high error bounds and high uncertainty.

Quality of all four indicators were poor for Victoria. Therefore, based on the current method, it was not possible to produce reliable life expectancy estimates for Victoria.

South Australia:

1. of the 38 age/sex combinations, 13 had less than three deaths in 2020–2022
2. difference in classification of Indigenous status in Census and deaths (27%)
3. death Identification rate (84%)
4. 95% confidence intervals for propensity was much wider than those jurisdictions that life expectancies are released for (that is, New South Wales, Queensland, Western Australia and the Northern Territory).

Quality of indicators 1 and 4 were poor for South Australia. Therefore, based on the current method, it was not possible to produce reliable life expectancy estimates for South Australia.

The life tables in this release were based on the average number of Aboriginal and Torres Strait Islander deaths registered in 2020, 2021 and 2022, and adjusted for differences in classification of Indigenous status.

Death records are collected as administrative data from the Registrars of Births, Deaths and Marriages in each state and territory. For all states and territories, death registration forms (DRFs) use a standard question to collect information about a person's Aboriginal and/or Torres Strait Islander origin. Medical Certificate of Cause of Death (MCCD) and coronial information was used as a secondary source for determining the Indigenous status of the deceased for all states and territories with the exception of Victoria. Use of this additional source has led to improved recording of Indigenous status. For more information see Causes of Death, Australia methodology.

The population used in this release is the final Aboriginal and Torres Strait Islander population estimates for 30 June 2021 based on the 2021 Census. For information on the method of calculating the Aboriginal and Torres Strait Islander estimated resident populations, see Technical note - Methodology used in Aboriginal and Torres Strait Islander population estimates.

Adjustment factors are the expected number of deaths divided by reported number of deaths. This was applied to the average of three years of registered deaths to account for underreporting or overreporting of Indigenous status in death records. They are the inverse of the identification rate.

• An adjustment factor of one indicates that there is no difference in the ratios of identification between data sources.
• An adjustment factor of less than one indicates that expected deaths using PES scale are lower than reported deaths and so deaths should be adjusted downwards.
• An adjustment factor of greater than one indicates that expected deaths using PES scale are higher than reported deaths and so deaths should be adjusted upwards.

The complete enumeration of a population or groups at a point in time with respect to well-defined characteristics (for example, Population, Manufacturing, etc.). In this release the word "Census" refers to the ABS Census of Population and Housing.

A confidence interval is a range of values that has been estimated to include the true population count for a specified level of confidence. A common confidence interval used in statistics is the 95% confidence interval. When an estimate has a 'normal probability distribution', the 95% confidence interval is constructed using approximately two (1.96) standard errors either side of the estimate.

Death is the permanent disappearance of all evidence of life after birth has taken place. The definition excludes all deaths prior to live birth. For the purposes of the Deaths collection compiled by the ABS, a death refers to any death which occurs in, or en route to Australia and is registered with a state or territory Registry of Births, Deaths and Marriages.

The official measure of the population of Australia is based on the concept of usual residence. It refers to all people, regardless of nationality, citizenship or legal status, who usually live in Australia, with the exception of foreign diplomatic personnel and their families. It includes usual residents who are overseas for less than 12 months over a 16-month period. It excludes overseas visitors who are in Australia for less than 12 months over a 16-month period.

Estimates of the Australian resident population are generated on a quarterly basis by adding natural increase (the excess of births over deaths) and net overseas migration (NOM) occurring during the period to the population at the beginning of each period.

Identification rate is the reported number of deaths divided by expected number of deaths. This is used to understand underreporting or overreporting of each Indigenous status in death records. It is the inverse of the adjustment factor.

A statistical process for predicting values where no response was provided to a question and a response could not be derived.

The Index of Relative Socio-economic Disadvantage (IRSD) used in this release was developed by the ABS and ranks areas in Australia from most disadvantaged to least disadvantaged using variables from the 2021 Census. It is one of four indexes published by the ABS, collectively known as Socio-economic Indexes for Areas (SEIFA).

In the IRSD index, a low score indicates a high proportion of relatively disadvantaged people in an area. A high score suggests that an area has a relatively low incidence of disadvantage, but it cannot be concluded that it has a large proportion of relatively advantaged people, as there are no variables in the index to indicate this.

Due to insufficient data for some areas, particularly at younger ages, the ABS has produced life tables for four categories of disadvantage. These are:

• Most Disadvantaged 20%
• Second Most Disadvantaged 20%
• Middle 20%
• Least Disadvantaged 40%.

For more information see Socio-economic Indexes for Areas (SEIFA), Australia, 2021.

In the Census, the method for collecting information about people who identified as being of Aboriginal and/or Torres Strait Islander origin was through self-identification. The following question was included in the Census and PES:

This question may be completed by the individual or by another member of the household on their behalf. For more information see Understanding change in counts of Aboriginal and Torres Strait Islander Australians: Census, 2021.

The average number of additional years a person of a given age and sex might expect to live if the age-specific death rates of each period continued throughout their lifetime.

Life expectancy summarises the mortality experience of a population by measuring how long, on average, a group of people born in the same year would be expected to live, if current death rates at each age remained the same. Life expectancy can be calculated for any age using life tables (a table which shows the probability of dying at each age), though life expectancy at birth is what most people focus on. Life expectancy at birth is the average number of years that a group of newborn babies would be expected to live if current death rates remain unchanged. Since death rates in the population will change during a person’s lifetime and a person may die at an earlier or later age, life expectancy is only a summary indicator for a population, rather than an exact measure of how long individuals will actually live.

A tabular, numeric representation of mortality and survivorship of a cohort of births at each age of life. The conventional life table is based on the assumption that as the cohort passes through life it experiences mortality at each age in accordance with a predetermined pattern of probabilities of dying (qx) which do not change from year to year. The life table is a hypothetical model of mortality, and even though it is usually based upon death rates from a real population during a particular period of time, it does not describe the real probabilities of dying which characterise a cohort as it ages.

Life tables may be complete or abridged, depending on the age interval used in their compilation. Complete life tables such as those for the Australian population contain data by single years of age. Abridged life tables such as those for the Aboriginal and Torres Strait Islander population contain data for five-year age groups. Due to differences in qx patterns between males and females at different ages, life tables are generally constructed separately for each sex.

The probability of dying (qx), is the main function of the life table, all other functions are derived from it. The life tables presented in this release contain four columns of interrelated information. These functions are:

• lx - the number of persons surviving to exact age x
• qx - the proportion of persons dying between exact age x and exact age x+n (where n is the width of the age interval). It is the probability of dying, from which other functions of the life table are derived
• Lx - the number of person years lived within the age interval x to x+n (where n is the width of the age interval), and
• ex - life expectancy at exact age x.

A linked record is a single record in one dataset that has been matched with a Census record. This was done using name and address on both datasets, and matching records using a probabilistic approach. If a match cannot be made, the record remains unlinked. These are referred to as unlinked records or records not linked.

PES and Census were also linked using clerical linking following the probabilistic approach of automated data linking. A team of coders manually confirmed or rejected candidate links provided by automated data linking using responses to name, sex, date of birth, age, marital status and more on both the PES and Census forms. The two datasets were linked in a way that was independent of reported Indigenous status so that any future analysis will not be affected by bias introduced in the linking process. For this reason, Indigenous status was not used as a linking variable. The coders search for people on Census forms at alternative addresses provided by the PES respondent, or in surrounding areas. For more information on PES and Census linking see 2021 Census overcount and undercount methodology, 2021 .

Linking death records with the Census enables a direct comparison of Indigenous status recorded on the death record and the Census, as they can be different across the two datasets or missing.

There were two main reasons why death records could not be linked to a Census record:

• Records belonging to the same individual were present in the death records and Census datasets but these records failed to be linked because they contained missing or inconsistent information; or
• A link was not possible because there was no Census record corresponding to the death record as the person was missed from the Census. Proximity of death to Census night was a significant factor in the ability for a link to be achieved.

For more information on death records and Census linking see 3302.0.55.004 - Linking Death registrations to the 2016 Census, 2016–17.

Excess of births over deaths.

The difference between the actual Census count (including imputations) and an estimate of the number of people who should have been counted in the Census. This estimate was based on the Post Enumeration Survey (PES) conducted after each Census. For a category of persons (e.g. based on age, sex and state of usual residence), net undercount was the resultant of Census undercount, overcount and imputation error.

Life tables for the non-Indigenous population were produced to compare life expectancy at birth and other ages between the Aboriginal and Torres Strait Islander population and non-Indigenous population of Australia.

Numbers of non-Indigenous deaths were obtained by subtracting the adjusted average number of Aboriginal and Torres Strait Islander deaths in 2020, 2021 and 2022 from the average number of total deaths registered in 2020, 2021 and 2022. The difference was the average annual number of non-Indigenous deaths.

Final estimates of the non-Indigenous population for 30 June 2021 from Estimates of Aboriginal and Torres Strait Islander Australians, 30 June 2021 were used as denominators in the calculation of age-specific death rates for the non-Indigenous population, and life tables derived from these.

Error which arises from inaccuracies in collecting, recording and processing the data. Every effort is made to minimise non-sampling error by the careful design of questionnaires, intensive training and supervision of interviewers, and efficient data processing procedures. Non-sampling error also arises because information cannot be obtained from all people selected in the collection.

Following the 1992 amendments to the Acts Interpretation Act to include the Indian Ocean Territories of Christmas Island and the Cocos (Keeling) Islands as part of geographic Australia, another category at the state and territory level was created, known as Other Territories. From 1 July 2016 the Australian Government assumed responsibility for Norfolk Island. Other Territories includes Jervis Bay Territory, previously included with the Australian Capital Territory, as well as Christmas Island, the Cocos (Keeling) Islands and Norfolk Island.

Person years are a function of life tables, denoted by Lx. It is calculated for each age interval (x to x+n, where n is the width of the age interval). It is the sum of:

• the number of years lived within the age interval by those surviving to x+n, and
• the number of years lived within the age interval by those who died during the interval (x, x+n).

A household survey conducted three to four weeks after the Census. The PES allows the ABS to estimate the number of people who should have been counted in the Census compared to the number who were. Results from PES contribute to a more accurate calculation of the estimated resident population (ERP) for Australia and the states and territories which is then backdated to 30 June of the Census year. In PES, the Indigenous status of each person is recorded (that is, there are no missing or ‘not stated’ responses).

Probability of dying, denoted by qx, is the probability that if a person has reached a certain age, what is the probability they die before they reach a given older age. This is also referred to as the mortality rate.

The probability of dying (qx), is the main function of the life table, all other functions are derived from it. The probability of dying is calculated using age-specific death rates for each age group and sex. Together the probability of dying and the probability of surviving each age group equal 1.

It is not the same as age-specific death rate, which is a rate of deaths amongst a population. See Age-specific death rates (ASDR).

A propensity is the probability of a particular outcome, or how likely it is that the outcome will occur, for a group of individuals with the same characteristics. The probabilities often vary between groups, so the propensities are described as the probability of an outcome given which group an individual belongs to. In this methodology we calculate propensities for groups of individuals that have the same, or different, Indigenous status on one dataset compared with another. For more information see adjustment factors.

While life expectancy at birth measures how long, on average, a group of people born in the same year can expect to live, it does not take into account how healthy they are expected to be throughout their life. Health adjusted life expectancy estimates are not regularly published, with the most recent data published by the Australian Institute of Health and Welfare in: Australian Burden of Disease Study: impact and causes of illness and death in Aboriginal and Torres Strait Islander people 2018.

The Australian Statistical Geography Standard (ASGS) Remoteness Structure divides Australia into five classes of remoteness which are characterised by a measure of relative geographic access to services. The five Remoteness Areas (RAs) are:

• Major Cities of Australia
• Inner Regional Australia
• Outer Regional Australia
• Remote Australia
• Very Remote Australia.

Due to insufficient data for some RAs, particularly at younger ages, the ABS has produced life tables for three categories of remoteness. These are:

• Major Cities
• Inner Regional and Outer Regional combined
• Remote and Very Remote combined.

While statistical data classed to this structure may be available by state/territory, characteristics of remoteness are determined in the context of Australia as a whole. Therefore, not all RAs are represented in all states and territories.

For more information and a map of the areas, see Australian Statistical Geography Standard (ASGS): Edition 3 - Remoteness Areas, July 2021.

Error which occurs in estimated quantities because a sample, rather than the entire population, is surveyed. One measure of the likely error in estimates resulting from not including all people in the survey, is given by the standard error.

Socio-Economic Indexes for Areas (SEIFA) is a product developed by the ABS that ranks areas in Australia according to relative socio-economic advantage and disadvantage. The indexes are based on information from the five-yearly Census.

SEIFA 2021 is the latest version of this product and consists of four indexes:

• The Index of Relative Socio-economic Advantage and Disadvantage (IRSAD)
• The Index of Relative Socio-economic Disadvantage (IRSD)
• The Index of Education and Occupation (IEO)
• The Index of Economic Resources (IER). 

Each index is a summary of a different subset of Census variables and focuses on a different aspect of socio-economic advantage and disadvantage.

For more information see Socio-Economic Indexes for Areas (SEIFA), Australia, 2021.

A measure of the spread of differences between the true population value and estimates from repeated samples. There are about two chances in three that a sample estimate from repeated samples will differ by less than one standard error from the true population value, and about nineteen chances in twenty that the difference will be less than two standard errors.

Survivorship is the measure within a life table of how many of the hypothetical 100,000 live birth cohort is still alive at the start of each age group. Together the probability of surviving and the probability of dying in an age group equal 1. See Life table.

lx - the number of persons surviving to exact age x

Some life tables include px, the probability of surviving through an age group. This is calculated as 1-qx. Qx is the probability of dying in an age group. See Probability of dying.

The intercensal growth in the Indigenous population counts that cannot be fully explained by births, deaths and migration.

In this publication, the year after Census refers to the period after Census night 2021, between 10 Aug 2021 and 28 Sep 2022. Deaths that occurred in this period are linked to 2021 Census records and used to calculate propensities. Also referred to as Census Year +1 or year after Census night.

The year the death occurred.

The year the death was registered.

To calculate estimates of life expectancy using direct methods, it was important to ensure that records were classified as representing Aboriginal and/or Torres Strait Islander peoples in a consistent manner in both the numerator (deaths) and the denominator (population). 

The Indigenous Mortality Study was first conducted as part of the ABS Census Data Enhancement (CDE) project for the 2006 Census. For more information on the CDE project, see the paper Linking Death Registrations to the 2016 Census.

In the absence of any unique identifier in the Census and deaths datasets, linking was performed using probabilistic methods. Three groups of variables, name (first name and surname), personal characteristics (date of birth, age, sex, place of birth, year of arrival and marital status), and geographic information (street number, street name, suburb, mesh block and postcode) were used to link death records to Census records. Variables common to both datasets were standardised to ensure consistent coding and formatting. The two datasets were linked in a way that was independent of reported Indigenous status so that any future analysis would not be affected by bias introduced in the linking process. For this reason, Indigenous status was not used as a linking variable.

The number of Census and death records eligible for linking and the percentage of death records linked are presented in Table 1. The linking process used:

• 202,735 death records and
• 25,422,788 Census records.

These Census counts are different from the estimated resident population, as discussed in Estimates of Aboriginal and Torres Strait Islander Australians, 30 June 2021.

• There were 202,735 death records
• 175,111 (86.4%) records were linked to one of 25,422,788 eligible Census records
• There were 4,754 Aboriginal and Torres Strait Islander death records
• 3,115 (65.5%) were linked.

This appendix describes the basis for estimating the sampling error of estimates of Aboriginal and Torres Strait Islander life expectancies at birth, and for estimating the sensitivity error of these life expectancy estimates from the assumptions made during calculation.

Broadly, a process of replication was used, where the inputs to the Aboriginal and Torres Strait Islander life expectancy calculations were replicated multiple times, based on sampling error information obtained from the Post Enumeration Survey (PES) and on plausible deviations from the assumptions. Variation among the replicate estimates of Aboriginal and Torres Strait Islander life expectancies, corresponding to replicate estimates from PES, gave estimates of the sampling error of life expectancy. The sensitivity error was measured from additional variation related to deviations from the method's assumptions.

The primary inputs to Aboriginal and Torres Strait Islander life expectancies were:

• Aboriginal and Torres Strait Islander estimated resident population (ERP) at 30 June 2021 (mid-point of the period from 2020 to 2022) by five-year age groups
• Average number of Aboriginal and Torres Strait Islander deaths over the period from 2020 to 2022, adjusted for differences in reported Indigenous status.

The method of calculating total Aboriginal and Torres Strait Islander ERP is described in Estimates of Aboriginal and Torres Strait Islander Australians, 30 June 2021.

The sampling error in PES generates sampling error in Aboriginal and Torres Strait Islander life expectancies. Standard errors (SEs) for life expectancies were calculated by first replicating PES inputs (based on estimated PES sampling error) 200 times and second, calculating the variance of the output from these replications. There were two kinds of PES input: population estimates (ERP) and identification rates.

• ERP: Estimates of the total and Aboriginal and Torres Strait Islander populations incorporate components of PES undercount adjustment.
• Identification rates: These were calculated from a probabilistic linking of death records (between 10 August 2021 and 28 September 2022) to the 2021 Census, and then to PES data. Identification rates were used to adjust for differences in reporting Indigenous status. The estimated identification rates will have sampling error as they use PES sample data.

The other data inputs without PES (for example, unlinked deaths and Census data) did not have associated sampling error, as they were not from a sample survey. Any error associated with these inputs was covered under the sensitivity error.

Note that PES data are weighted to estimate the propensities but unweighted data were used to illustrate the methodology in Appendix: Census and deaths data linkage results.

In calculating and applying the identification rate the following assumptions were made:

• identification rates for unlinked deaths were assumed to be the same as those for linked deaths.
• identification rates for the entire period from 2020 to 2022 were assumed to be the same as those observed during the year after Census (August 2021 to September 2022).
• identification rates were assumed to apply uniformly across sex and within each broad age group.
• propensities for identifying as Aboriginal and/or Torres Strait Islander in the Census and PES were assumed to be applicable to death records, even with different age distributions and methods of collection for deaths.
• Census/PES non-Indigenous and not stated propensities for Australia were assumed to apply uniformly across regions (state/territory or Remoteness Area or IRSD group).

There was no direct information on possible deviation from the first two assumptions, but deviations within plus or minus five percentage points of identification rate for each state/territory were used for the first assumption, and within plus or minus two percentage points for each sex and state/territory for the second assumption. Data from the linked file and PES were analysed to give ranges of deviations in identification rates for the last three assumptions.

The 95% confidence intervals for Aboriginal and Torres Strait Islander life expectancy at birth, 2020 to 2022 estimates from the sensitivity analysis and sample error are given in Tables 1, 2 and 3 below.

The sensitivity and sample errors for Aboriginal and Torres Strait Islander death identification rates from the sensitivity analysis and sample error are given in Tables 4, 5 and 6 below.