One important dimension of the capacity of Indigenous people and families to pay for health expenditure is family or household income. In order to calculate the resources available for improving health, one must appreciate the overall demands on resources within a family or household. There is no consensus on how this should be done, and several standard techniques exist to adjust family income to allow for the different number and characteristics of family members—that is, to apply an equivalence scale.^[4] Note that while equivalence scales are standard tools in poverty analysis, they will be used in this report to provide an indication of the overall access to resources within families, some of which could potentially be spent on health services.

As indicated above, the NHS provides income data adjusted using the ABS’s version of the simplified Henderson equivalence scales. However, this is only one of several scales available and, as noted, there is an ongoing controversy about the precise specification of equivalence scales. This revolves around the nature and extent of economies of scale in families or households—the smaller the proportion of expenditure on items which display economies of scale, the more justifiable it is simply to divide family or household income by the number of people it supports (Guobao, Richardson & Travers 1996). When income levels are very low, a high proportion of expenditure is on food, basic clothing, cooking fuel and certain health expenditures. Given that each of these varies directly, and quite closely, with the number of people in the family, it may make it appropriate to give each person a similar weight by focusing on per capita income. In contrast, where so-called ‘public goods’ (i.e. where a certain expenditure improves the well-being of all residents and not only the person consuming the resources) are important, such as in various categories of health expenditure, more account needs to be taken of potential economies of scale implicit in the equivalence scales. At the extreme, raw income measures implicitly assume that extra family members cost no more to maintain than the first person. While this assumption is obviously untenable, it provides a useful bound on possible assumptions about economies of scale.

The point to note from this sensitivity analysis is that the relationship between health expenditure and income is likely to be distorted through measures of income which do not adjust for the number and age of people living off that income.^[5] The per capita equivalence scales overestimate the needs of larger families in comparison with smaller families (De Vos & Zaidi 1997). In contrast, the use of raw income underestimates the needs of such families. The equivalence scales used in this paper cover the majority of possible assumptions about household costs, ranging from there being no extra cost to additional persons living in the family to there being no economies of scale in people living together. Given the preponderance of larger families in the Indigenous population the analysis of Indigenous health expenditure may be particularly sensitive to the type of income measure used.

Another tension implicit in choosing the appropriate equivalence scale for Indigenous income units is that the definition of the appropriate unit of analysis is not obvious in the Indigenous context (Altman & Hunter 1998). The widely used Henderson equivalence scales may be appropriate for a nuclear family, but it is more difficult to rationalise their use when Indigenous households can be characterised as having: compositional complexity; porous social boundaries and large size; extended families resident in one or more dwellings; households being subject to considerable fluctuation; and small, multi-generational core(s), dissolving and reforming in developmental cycles (Altman et al. 1997). Hunter and Smith (2000) have argued that focusing on households, rather than families, makes comparisons between Indigenous and non-Indigenous populations particularly problematic, especially when using the standard ABS definitions. Because of the conceptual difficulties in measuring income in a cross-cultural context, this paper uses a variety of equivalence scales on family income to capture the likely sensitivity of results to the underlying assumptions about economies of scale and access to resources.^[6]

The income measures are calculated for income units (i.e. families) using four equivalence scales: raw income, the Henderson scale, the new OECD scale and per capita income.^[7] Raw income is simply the sum of income of family members. The other income measures adjust for the size and composition of families by dividing this raw income by their respective equivalence scale.

While Henderson’s scale has been the standard measure for equivalent income in Australia since the mid-1970s, there is increasing criticism of the robustness of the resulting estimates (Henderson 1975; Saunders 1994; Travers & Richardson 1993). Accordingly, two extra measures of equivalent income are included to explore the feasible range of access to resources. However, Henderson’s scale does have the advantage that it is the only scale to attempt to control for extra costs incurred by working or looking for a job. This adjustment is likely to be particularly important when comparing estimates of Indigenous and other Australians, given the enormous disparity in employment rates between these groups (Taylor & Hunter 1998).

Another equivalence scale widely used in international studies of poverty is the OECD scale. This paper uses the new or modified OECD scale, which gives a weight of one to the first adult, 0.5 to the second and subsequent adults, and 0.3 to all dependants (see De Vos & Zaidi 1997 for further details of the history of the OECD equivalence scales).

The last income measure used is per capita family income. This is calculated by dividing the raw income by the number of people in a family. The advantage of using these four income measures is that they cover the range of possibilities of economies of scale and access to resources. As discussed above, raw income and per capita income provide the extreme bounds of possible assumptions, with the Henderson and new OECD measures falling somewhere within these bounds. While the Henderson and new OECD scales probably provide more feasible estimates of access to resources, the sensitivity analysis needs to test whether our results are robust to all possible measures. Note that Fig. 3.1, and all subsequent analysis, reports the equivalent income measures in descending order of implicit economies of scale: raw income, the new OECD scale, the Henderson scales and per capita income.^[8]

Income quintiles for these four different measures of income were estimated from the 1995 NHS separately for the Indigenous and non-Indigenous components of the population (Fig. 3.1). Each estimate is ranked according to its place in the overall distribution of the respective measures of equivalent income in the 1995 NHS. That is, the income quintiles used in this paper are measured for the Australian population using NHS 1995 data. Accordingly, the non-Indigenous distribution, which dominates the overall income distribution, is even, with 20 per cent being in each quintile.

In line with Deeble et al. (1998), Fig. 3.1 illustrates that Indigenous people are disproportionately concentrated in the low income groups. As this earlier study only reported the distribution of equivalent income using a simplified Henderson scale, it is useful to compare this with other income distributions. One obvious point to make is that the per capita measure of equivalent income is even more concentrated in the low income group (at least, the bottom quintile). For example, per capita income is about 10 percentage points more likely to classify Indigenous families in the lowest quintile than the Henderson measure. On the other side, Henderson classifies over 10 per cent more of the Indigenous population in the bottom quintile than does raw income. Notwithstanding this, the overall shape of the distribution is similar, with most of the differences occurring in the first and second quintiles. The top two quintiles have very similar numbers of Indigenous families in all four income distributions.

Even though some of the overall income distributions in Fig. 3.1 do not differ much for the various measures of equivalent income, there are substantial reclassifications of families between the respective scales. Large families are more likely to be in the high quintiles of raw income, irrespective of living circumstances. While such families will tend to be reclassified in the lower income groups with the other equivalence scales (especially the per capita measures), other family types will be reclassified into higher income groups. The fact that Indigenous families are almost twice as likely to have a sole parent than other families with children complicates the comparisons between Indigenous and other Australian families (Daly & Smith 1998a; Daly & Smith 1998b). The extent of reclassification of family income depends crucially upon the number of children in the family and the assumption made about the relative costs of children and adults for the respective equivalence scales (Hunter, Kennedy & Smith 2001).

The age profiles of the various income groups can affect the interpretation of an income-based analysis.^[9] For example, if low income groups include disproportionate numbers of older people, who would be more likely to be sick irrespective of their income status, then improvements in health as one moves up the income distribution may be driven as much by demographic factors as by differential access to resources or information. A brief perusal of age profiles by income reveals that this is the case for almost all measures of equivalent income used in this report, especially for the non-Indigenous population. The exception is per capita income, where the bottom quintile group has relatively few people aged 55 years or more (four times fewer than in the second quintile). The likely reason for this is that per capita income tends to re-rank small families at the end of the lifecycle (i.e. where the children have left home) into high-income quintiles compared with large families with many dependants. Whatever the reason, the fact that the age profile differs across income measures means that if the following analysis points to consistent differentials, irrespective of the equivalence scale, then it is possible to rule out that the analysis is driven by demographic factors.

Detailed analysis of the NHS income data indicates that there is substantial re-ranking of families or income units across income quintiles, with as many as one-third of families changing income group when different equivalence scales are used (Hunter, Kennedy & Smith 2001). Given the substantial reclassification of income groups for both Indigenous and non-Indigenous populations, it would be surprising if the analysis of health expenditure was not sensitive to the choice of equivalence scales. Exploration of these effects provides for a more sophisticated treatment of income than was possible in Deeble et al. (1998) and yields greater insight into the relationship between income, health status and expenditure.