Widening ethnic mortality disparities in New Zealand 1981–99
Introduction
New Zealand life expectancy is mid-range for OECD or developed countries (78.9 years in 2002, sexes combined (World Health Organization, 2003)). This figure, however, fails to disclose marked social group differences in life expectancy. For example, males and females living in socio-economically deprived neighbourhoods during 1995–97 had life expectancy of 9 and 7 years less than males and females living in non-deprived neighbourhoods, respectively (Salmond & Crampton, 2000). Likewise, Mäori (the indigenous population of New Zealand; approximately 15% of the population in 2001) and Pacific people (people born in the Pacific Islands or their descendants; approximately 6% of population) have lower life expectancy than the majority non-Mäori non-Pacific (nMnP) group (Ministry of Health, 1999a). However, trends in life expectancy and mortality by ethnicity in New Zealand are not well documented and understood. This paper documents trends in life expectancy and mortality by ethnicity and summarises a body of work undertaken and disseminated within New Zealand (Ajwani, Blakely, Robson, Atkinson, & Kiro, 2004; Ajwani et al., 2003; Blakely, Ajwani, Robson, Tobias, & Bonne, 2004). Why is this work of interest to an international audience?
First, New Zealand is a unique ‘case study’ due to a mix of social, cultural and macroeconomic factors. In the 1950s and 1960s, New Zealand society fared well with one of the highest living standards and GDPs in the world. New Zealand's agricultural economy was strong, in large part due to a privileged position as supplier of butter, meat and wool to Britain, lasting until Britain entered the European Economic Community in 1973 (Belich, 2001). A strong and growing economy underpinned a strong welfare society with reasonably good access to health and other services. Mäori moved rapidly from a rural to urbanised population, responding to the need in major centres for workers in manufacturing and other industrial sectors. In 1946, 74% of Māori were rural dwellers but by 1976, 76% of Māori lived in urban areas having moved there for employment (Pool, 1991). In contrast, European settlers had become predominantly city dwellers by 1911 (King, 2003).
New Zealand society changed dramatically in the 1980s and 1990s, with 1984 often identified as the major turning point following a decade or so of falling terms of trade and double-digit inflation. In response, from 1984 to the early 1990s, New Zealand underwent major social and economic changes including a substantially flattened tax system, fully targeted income support, a regressive consumption tax, market rentals for housing, privatisation of major utilities, user charges for health, education and other government services, and a restructured labour market designed to facilitate ‘flexibility’ (Belich, 2001; Boston, Dalziel, & John, 1999; Cheyne, O’Brien, & Belgrave, 1997). Migration of Pacific people (e.g. Samoan, Cook Island) to New Zealand occurred predominantly in the 1950s to mid-1970s, driven initially by the employment opportunities (Cook, Didham, & Khawaja, 1999).The social and macroeconomic changes weighed particularly heavily on Mäori (and Pacific people), with inequalities between Mäori and non-Mäori widening in employment status, education, income and housing (Mowbray, 2001; Te Puni Kokiri, 2000). It has previously been speculated that this economic restructuring may have been a contributing reason to New Zealand's life expectancy falling behind that of its neighbour Australia from the 1970s onwards, and further that the health impacts of the structural reforms on mortality may have been differential by ethnicity (Brown, 1999; O’Donoghue, Howden-Chapman, & Woodward, 2000).
The second reason why this study is of international interest is a methodological one. It has been known for some years in New Zealand that mortality data underestimate the number of Mäori and Pacific deaths relative to that collected by the census (Graham, Jackson, Beaglehole, & de Boer, 1989; Pomare et al., 1995). The reason for this undercount is that, prior to September 1995, each decedent could be identified as either Mäori or Pacific based on ‘degree of Mäori (or Pacific) blood’ of the deceased's parents. Using this information, ethnicity was coded as just one of three options—Mäori, Pacific, nMnP—using a rule of ‘Māori (or Pacific) blood greater than half, otherwise nMnP’. This information was often not collected by the undertaker, and missing data were coded as nMnP immediately introducing a systematic bias to undercount Mäori and Pacific deaths. After September 1995, mortality data used an approximation to the 1996 census ethnicity question—and it was compulsory.
The denominator Census data have from 1986 onwards used multiple self-identified ethnic groups (i.e. it was not based on biological origin). (The 1981 census used multiple self-identified ethnic origin, specified in fractions.) These different ethnicity collection systems gave rise to numerator–denominator bias when calculating ethnic mortality rates: the mortality data numerators are not consistent with the census data denominators. However, the magnitude of this bias, and how it varied over time, was unknown. This is the major reason why trends in mortality by ethnic group in New Zealand are poorly understood.
The advent of the New Zealand Census-Mortality Study (NZCMS) (Blakely, Salmond, & Woodward, 2000), with its linkage of census and mortality data, has enabled a direct comparison of ethnicity recording on census and mortality data during most of the 1980s and 1990s. This comparison allows the calculation of adjustment factors for numerator–denominator bias that can be applied to ethnic mortality rates. The potential for such numerator–denominator bias has been recognised previously in the United States (Rosenberg et al., 1999; Sorlie, Rogot, & Johnson, 1992). In the US Black and White mortality rates were found to be reasonably accurate, but mortality rates for Native Americans and Hispanics during the 1980s were underestimated by 27% and 7%, respectively, due to under-reporting of these groups on mortality data. The introduction of multiple race categories in the 2000 US census is likely to mean that significant numerator–denominator bias now also exists for Black mortality rates (Ingram et al., 2003). It is likely that numerator–denominator bias by ethnicity also exists in other countries (e.g. Australia for Aboriginal and Torres Strait Island people and Canada for First Nations people). Our New Zealand experience may stimulate similar scrutiny in these countries.
The objectives, therefore, of this paper are:
- 1.
To determine the extent of undercounting of Mäori and Pacific peoples’ deaths in New Zealand during the 1980s and 1990s.
- 2.
To determine corrected ethnic mortality trends during the 1980s and 1990s by sex, age group and cause of death.
- 3.
To determine corrected ethnic life expectancy trends during the 1980s and 1990s, and the contribution of age at death and cause of death to ethnic disparities.
Section snippets
Mortality data: 1980–1999
Mortality data were provided by the New Zealand Health Information Services (NZHIS) for the years 1980–1999 by year of registration of death. Years were grouped into four periods: 1980–84, 1985–89, 1990–95 and 1996–99. The third period is of 6 years duration and the fourth period of 4 years duration to reflect the change in the collection of ethnicity in mortality data from September 1995.
Census data
For each of the above four periods, 1981, 1986, 1991 and 1996 census data by strata of sex, age and
Numerator–denominator bias
The previous convention and best practice for calculating ethnic mortality rates in New Zealand had been to use the following denominator data: ‘half or more blood quantum’ counts from the 1981 census, ‘sole’ ethnicity counts from the 1986 and 1991 censuses, and ‘prioritised’ ethnicity counts from the 1996 census. Regarding numerator data, there was no choice up to September 1995 (i.e. single options allowed only, and missing treated as nMnP), and prioritised counts have been used since 1996.
Discussion
There are two major foci of this paper: adjustment for numerator–denominator bias, and the substantive issue of varying trends in mortality rates and life expectancy by ethnic group in New Zealand.
Conclusion
We found large biases in mortality statistics due to undercounting of Mäori and Pacific deaths during the 1980s and early 1990s in New Zealand. We urge researchers in other countries to consider the possibility of numerator–denominator bias affecting ethnic mortality rates in their country. Having corrected for this numerator–denominator bias, we found marked differences in ethnic mortality rates and trends in these rates. Following three decades of converging life expectancy trends in New
Acknowledgements
The New Zealand Census Mortality Study (NZCMS) is conducted in collaboration with Statistics New Zealand and within the confines of the Statistics Act 1975. The Health Research Council of New Zealand and the Ministry of Health provided funding. We wish to thank the following people who commented on this research: June Atkinson, Cindy Kiro, Jackie Fawcett, Papaarangi Reid, Andrew Sporle, and the many Pacific and Mäori people and other colleagues that were consulted. This paper is published with
References (70)
- et al.
Social inequality and ethnic differences in smoking in New Zealand
Social Science & Medicine
(2004) - et al.
Anonymous linkage of New Zealand mortality and Census data
Australian and New Zealand Journal of Public Health
(2000) Epidemiology and the web of causation: Has anyone seen the spider?
Social Science & Medicine
(1994)- et al.
Changes in life expectancy in Russia in the mid-1990s
Lancet
(2001) - et al.
Explaining trends in inequities: Evidence from Brazilian child health studies
Lancet
(2000) - et al.
Age standardization of rates: A new WHO standard
(2000) - Ajwani, S., Blakely, T., Robson, B., Atkinson, J., Fawcett, J., & Kiro, K. (2002). Unlocking the numerator–denominator...
- et al.
Unlocking the numerator–denominator bias III: Adjustment ratios by ethnicity for 1981–1999 mortality data. The New Zealand Census-Mortality Study
New Zealand Medical Journal
(2004) - et al.
Decades of disparity: Ethnic mortality trends in New Zealand 1980–1999
(2003) - et al.
Maori disadvantage in the labour market
(2002)