J Pop Research (2012) 29:283 287 DOI 10.1007/s12546-012-9096-3 Response: ABS s comments on Estimating Indigenous life expectancy: pitfalls with consequences M. Shahidullah Published online: 18 August 2012 Ó Springer Science & Business Media B.V. 2012 This is an Australian Bureau of Statistics (ABS) response to Madden et al. s paper entitled Estimating Indigenous life expectancy: pitfalls with consequences, which is published in this issue of the Journal of Population Research. This paper is a critique of the method used by the ABS for compiling life tables for Aboriginal and Torres Strait Islander Australians. The paper claims that the ABS methods have understated Indigenous deaths and so overstated life expectancy. The purpose of this rejoinder is to ensure that readers are aware of the ABS response to the points raised in the paper. The Council of Australian Governments (COAG) committed themselves collectively to overcoming the disadvantage experienced by Aboriginal and Torres Strait Islander Australians and set a number of high-level targets to achieve this commitment. One of the targets is to close the life expectancy gap between Aboriginal and Torres Strait Islander and non-indigenous Australians (Steering Committee for the Review of Government Service Provision 2009). The Ministerial Council for Aboriginal and Torres Strait Islander Affairs agreed to support jurisdictional improvement in the quality of Aboriginal and Torres Strait Islander demographic data, with priority given to improve the compilation of Aboriginal and Torres Strait Islander life tables and life expectancy estimates. Therefore, a critical examination of the method of compiling life expectancy estimates for Aboriginal and Torres Strait Islander Australians is welcome. The main criticisms of the Madden et al. paper are related to: 1. Apparent widespread rejections of the ABS methods; 2. ABS s handling of Indigenous classification in a consistent manner in the numerator and the denominator rather than using alternative estimates; M. Shahidullah (&) Demography Section, Australian Bureau of Statistics, Belconnen, ACT, Australia e-mail: m.shahidullah@abs.gov.au
284 M. Shahidullah 3. ABS not having used the ever-indigenous rule to identify Indigenous deaths in the Census Data Enhancement Indigenous Mortality Quality Study; 4. ABS having extended its methodology for state/nt estimates to calculate estimates for sub-state geographies for projecting Indigenous population; 5. Coverage of registered Indigenous deaths for WA and the NT; and 6. ABS s intention of using the same consistent methodology after the 2011 Census. ABS s comments on each of these points are provided below: 1. Apparent widespread rejection of the ABS methods. Madden et al. claim there was broad acceptance in principle of this proposal. But there was widespread rejection of the methods. This claim is not valid as the new ABS method for compiling Indigenous life tables received overwhelming support (i.e. 21 of the 25 submissions to ABS supported the method) from key government agencies and from academics with an interest in Indigenous mortality statistics. The ABS method also gained support from overseas experts. For example, Tony Blakely, the leader of the New Zealand Census-Mortality Study that conducted similar work over the last 10 years based on linked census-mortality data, strongly supported the ABS method (in communication with the ABS on 11 November 2008). Only a limited number of academics and researchers expressed concerns about the method for adjustment for under-coverage of Indigenous deaths in death registration. Academics and researchers who identified concerns about the method for adjustment suggested that ABS should consider: including deaths which were identified as Indigenous in either death registration or the Census (i.e. the ever- Indigenous rule) for deriving coverage estimates; incorporating preliminary results from the Australian Institute of Health and Welfare s data linkage study to derive life tables; and providing error bands to indicate the uncertainty associated with life expectancy at birth estimates. ABS welcomed these suggestions and incorporated them into the final life tables paper (ABS 2009), which included Indigenous life expectancy estimates using the following six alternative approaches for adjusting deaths data as input to compilation of life tables: using deaths identified as Indigenous in either the Census or death registrations in the ABS Census Data Enhancement (CDE) linked data only; using deaths identified as Indigenous in either the Census or death registrations in the CDE linked and unlinked data; using the Chandra Sekar and Deming Capture/Recapture approach; using Indigenous identification based on the 2006 Census; using CDE-adjusted Indigenous deaths, with further adjustments to the age and sex structure of deaths according to preliminary results from the Australian Institute of Health and Welfare Indigenous Mortality Data Linkage Study; and directly using deaths identified as Indigenous in the Australian Institute of Health and Welfare Indigenous Mortality Data Linkage Study. 2. ABS s handling of Indigenous classification in a consistent manner in the numerator and the denominator rather than using alternative estimates. Aboriginal and Torres Strait Islander deaths data in Australia and the Maori and Pacific Islander deaths data in New Zealand are both sourced from death registration
ABS comments on Estimating Indigenous life expectancy 285 forms usually completed by a funeral director. Death registration data have a common data quality issue, namely the underreporting and/or incorrect reporting of a person s Indigenous status on the death registration form. The New Zealand Census-Mortality Study team, led by Tony Blakely, linked the 1981, 1986, 1991 and 1996 Censuses to three years of subsequent mortality data to study Maori, Pacific, non-maori and non-pacific mortality in New Zealand. This team used the numerator and denominator consistently to derive adjustment factors for undercounting of Maori and Pacific deaths and overcounting of non-maori and non-pacific deaths in death registrations (Ajwani et al. 2003; Blakely et al. 2002a, b). To derive estimates of life expectancy, it is important to ensure that the classifications of records as Indigenous occurs in a consistent manner in both the numerator (deaths) and the denominator (population). ABS analysis suggests that using a consistent identifier reduces the error in both mortality rates and life expectancy estimates. Therefore, like the New Zealand study, the ABS study also used the numerator and denominator consistently in deriving adjustment factors for undercoverage. 3. ABS not having used the ever-indigenous rule to identify Indigenous deaths in the Census Data Enhancement Indigenous Mortality Quality Study. Madden et al. mentioned reporting as Indigenous in one collection is sufficient to count an individual as Indigenous in both collections. They seem to be overly optimistic about the usefulness and validity of the ever-indigenous rule. Their views contradict the widespread concerns expressed by many stakeholders including some Australian State Health Departments that the ever-indigenous approach can lead to potentially spurious over-identification of Indigenous deaths. Owing to data collection, transcription or processing errors, a non-indigenous person may be incorrectly identified and recorded as Indigenous in any of the linked records, resulting in their being deemed Indigenous across all records. The authors referred to the Australian Institute of Health and Welfare (AIHW) study which compiled Indigenous life expectancy estimates by linking registered deaths with Indigenous death records from three alternative data sources. This study produced life expectancy at birth estimates of 66.8 years for Indigenous males and 72.9 years for Indigenous females for the period 2001 2006 (AIHW 2011). The ABS estimates were 67.2 and 72.9 years for Indigenous males and females respectively for the period 2005 2007. It is important to note that there were differences in methodologies and reference periods between the ABS and AIHW studies. However, despite these differences, results were remarkably similar: the AIHW male estimate was 0.4 years lower than the ABS male estimate, while the AIHW female estimate was exactly the same as the ABS female estimate. On 18 July 2012, AIHW released a report on Indigenous life expectancy estimates (AIHW 2012). These estimates were based on the ever-indigenous approach, and the report concluded There is still considerable work to be done before the AIHW can be confident in the results of such linkage work (AIHW 2012, p. 23). This suggests that even the authors themselves are quite uncertain about the usefulness of the ever-indigenous approach. Therefore, further study is needed to justify the merit of using this approach in compiling Indigenous life tables.
286 M. Shahidullah 4. ABS having extended its methodology for state/nt estimates to calculate estimates for sub-state geographies for projecting Indigenous population. There was, and still is, a high demand for Indigenous population projections at the Indigenous Regions and Remoteness Area levels. In order to project Indigenous population for these geographies, the ABS formulated mortality assumptions from the state/nt life expectancy estimates. In accordance with the usual, the ABS prepared a detailed paper on the proposed assumptions to be used for compiling the 2006 Census based Indigenous projections at the national, state and territory and the specific sub-state levels. ABS was transparent and open with the proposed assumptions and sought feedback from key stakeholders regarding the suitability and validity of the assumptions. Stakeholders commented that the proposed assumptions were sensible and supporting using these assumptions in projecting the Indigenous population. 5. Coverage of registered Indigenous deaths for WA and the NT. Madden et al. reasonably observed that In WA and the NT, there is the implication that some deaths reported as Indigenous are reported in error. Based on recent evidence, this statement is true particularly for WA. Investigation conducted by the WA Registrar of Births, Deaths and Marriages indicated that some non- Indigenous deaths were wrongly identified as Indigenous deaths in WA for 2007, 2008 and 2009. ABS discussed this issue with key stakeholders and users of Aboriginal and Torres Strait Islander deaths statistics. Following this discussion, ABS has not released WA Aboriginal and Torres Strait Islander deaths data for the years 2007, 2008 and 2009 in the 2010 issue of Deaths, Australia publication. The WA Registry corrected the data and resupplied the corrected data to ABS. These corrected data were then released by ABS in spreadsheets attached to Deaths, Australia, 2010 (ABS 2011) publication on 24 May 2012. 6. ABS s intention of using the same consistent methodology after the 2011 Census. Policy makers place considerable importance on producing statistics based on consistent methodology over time. To ensure that the Indigenous life tables for the periods 2005 2007 and 2010 2012 could be derived using the same method, COAG funded ABS to link the 2011 Census data to death registration data. The benefit of using the same method is that it will enable a valid comparison between the 2005 2007 and the 2010 2012 life tables and thus assist in monitoring whether the gap in life expectancy estimates between Indigenous and non-indigenous Australians has been widening or closing. Conclusion It is important to note that no method of compiling life tables for Indigenous Australians is perfect as all methods rely on some level of assumption. It is well established that the source data on Indigenous births and deaths can be subject to a
ABS comments on Estimating Indigenous life expectancy 287 number of quality limitations, such as late registration of births and deaths, underreporting and incorrect reporting of a person s Indigenous status on the registration form. The key priority for the ABS in compiling Indigenous life tables is to adopt a robust and defensible method which is fully transparent in outlining all contributing assumptions and known data quality issues. Additionally, ABS s role in informing evidence-based policy development means that consideration must be given to the needs of stakeholders in using the statistics produced. ABS recognizes that any method has limitations and will be based on certain assumptions which may introduce some bias. However, ABS undertook extensive sensitivity analysis on the life expectancy estimates it produced which suggested that such bias was minimal (ABS 2009, p.30). Acknowledgments I would like to thank Patrick Corr, Nicholas McTurk and Jill Charker of the Australian Bureau of Statistics for their useful comments and suggestions on this rejoinder. References Ajwani, S., Blakely, T., Robson, B., Atkinson, J., & Kiro, C. (2003). Unlocking the numeratordenominator bias III: Adjustment ratios by ethnicity for 1981 1999 mortality data. The New Zealand Census-Mortality Study. New Zealand Medical Journal, 116, 1 12. Australian Bureau of Statistics (ABS). (2009). Experimental life tables for Aboriginal and Torres Strait Islander Australians, 2005 2007. Cat. No. 3302.0.55.003, Canberra. Australian Bureau of Statistics (ABS). (2011). Deaths, Australia, 2010. Cat. No. 3302.0, Canberra. Australian Institute of Health and Welfare (AIHW). (2011). Life expectancy and mortality of Aboriginal and Torres Strait Islander people. Cat. No. IHW 51, Canberra. Australian Institute of Health and Welfare (AIHW) (2012). An enhanced mortality database for estimating Indigenous life expectancy: A feasibility study. Cat. No. IHW 75. Canberra. Blakely, T., Kiro, C., & Woodward, A. (2002a). Unlocking the numerator-denominator bias II: Adjustment to mortality rates by ethnicity and deprivation during 1991 94. The New Zealand Census-Mortality Study. New Zealand Medical Journal, 115, 43 48. Blakely, T., Robson, B., Atkinson, J., Sporle, A., & Kiro, C. (2002b). Unlocking the numeratordenominator bias I: Adjustment ratios by ethnicity for 1991 94 mortality data. The New Zealand Census-Mortality Study. New Zealand Medical Journal, 115, 39 43. Steering Committee for the Review of Government Service Provision. (2009). Overcoming indigenous disadvantage: Key indicators 2009. Canberra: Productivity Commission.