Appendix 1: Conceptual and technical issues
- Absolute and relative measures
- Disposable income, discretionary income and housing
- Equivalence scales
- Zero and very low incomes
There are a variety of technical and measurement issues that affect estimates of the level of poverty. These issues include the definition of disposable income, the establishment of equivalence scales between different types of households, the underreporting of household incomes, the method of sampling, the handling of missing income data, the assignment of weighting, the unit of analysis and whether to use weekly or annual income. In the Australian context, these issues have been discussed elsewhere (Greenwell, Lloyd & Harding 2001; Harding, Lloyd & Greenwell 2001; Johnson 1987, 1996). The purpose of this appendix is to provide an overview of issues relevant to the estimates presented in this paper.
Absolute and relative measures
Relative measures are the dominant approach to the study of income poverty in Australia and in other western countries. The poverty line is most often drawn at half the median disposable household income after adjusting for household size (Atkinson 1998; Brady 2003; Moller et al. 2003; Oxley et al. 1997). Less often, the poverty line has been drawn at half the mean disposable income (Harding, Lloyd & Greenwell 2001). Relative measures are easy to understand and especially well suited to cross-national studies. Unlike absolute measures, they do not require justifications for a particular basket of goods and services. Furthermore, they do not require updating.
A common criticism of the relative approach to income poverty is that it measures distribution rather than financial disadvantage (Saunders & Kayoko 2002). There is a sense that with relative measures, poverty will always be with us. It is often pointed out that if the real incomes of all households doubled over the next 10 years, relative poverty would remain the same. Mean-based measures are particularly sensitive to the distribution of income; if the real income of higher income households grew more strongly than that of other households, poverty would increase. Relative poverty can only be eliminated by making radical changes to the distribution of household income—not a feasible or desirable policy option—rather than substantially increasing the standard of living of the lowest income households. Another criticism of relative measures is their arbitrariness. There is no particular reason why 50 per cent was chosen for drawing the poverty line. For example, the Eurostat Task Force recommended it be drawn at 60 per cent of median income (Eurostat Task Force 1998). Furthermore, relative measures do not have the support of the Australian public that absolute measures do (Saunders 2004, p. 8). Finally, half the median or mean income measures give no indication of the standard of living. An income below half the median income provides a very different standard of living in Australia to that in other industrialised countries (Kangas & Ritakallio 2004b).
An alternative approach is to use absolute measures of poverty. Absolute measures are defined as having insufficient income to purchase the very basic physical necessities of life, such as food to eat.10 Absolute measures of poverty were the dominant approach for the first half of the 20th century (Saunders Bradshaw & Hirst, 2002). In Australia, the original HPL was defined in absolute terms as the basic wage plus child endowment for a family of four in the mid-1960s (Henderson, Harcourt & Harper 1970; Saunders 1998a).11 The income required for other family types was calculated from this benchmark. Also formulated in the 1960s was the official poverty line for the United States, which was defined as three times the cost of a basic food basket.12
The problem with absolute measures of income poverty is that very few citizens of industrialised societies live without shelter, running water or sufficient food, and so to define poverty in such absolute terms would be to define it out of existence. Furthermore, there is no community consensus on the minimum income required to live decently (Saunders 1998a). Similarly, there would be little agreement among experts on the selection and level of goods and services required to live in contemporary Australia.
There is an argument that distinguishing absolute and relative conceptions of poverty is a false dichotomy. The Henderson and official United States poverty lines first formulated in the 1960s have been continuously updated in an attempt to preserve their original levels. Sen's (1987) influential theoretical work on poverty combines absolute and relative approaches. Poverty is defined in terms of not having the capabilities to function in society at minimally acceptable levels. This includes adequate nourishment, clothing, shelter and access to health services as well as participation, or at least the capacity to choose to participate in a wide range of aspects of modern society. Nevertheless, making a distinction between absolute and relative approaches to poverty remains useful.
Estimates of the extent of poverty in Australia
Absolute and relative measures of income produce different estimates of the extent of poverty in Australia and are very sensitive to the criteria used. In the year 2000, 22 per cent of adults and children were below the HPL. Absolute measures based on half-mean and half-median disposable incomes indicate substantially lower levels of poverty, at 13 and 9 per cent respectively (using the same Henderson equivalences). The HPL also shows much higher rates of poverty among children at around 25 per cent, compared to 15 and 10 per cent respectively with the mean and median measures (Harding, Lloyd & Greenwell 2001, p. 35).
Different cut-off lines for the relative measures provide very different estimates of the extent of poverty in Australia. About 6 per cent of Australians are in poverty when the poverty line is defined at 40 per cent of the median equivalised disposable income; the proportion rises to about 20 per cent when the cut-off is 60 per cent (Förster 1994; Harding & Mitchell 1992).
Mean-based and median-based measures can lead to quite different conclusions on changes in poverty over time. According to the half-mean measure with Henderson equivalences (see below for a discussion of equivalence scales), poverty increased from 11.3 per cent in 1990 to 13.0 per cent in 2000. The half-median measure shows a much smaller increase from 8.2 to 8.7 per cent (Harding, Lloyd & Greenwell 2001, pp. 4–5). It is difficult to reconcile an increase in poverty during the late 1990s with increases in the real incomes of low-income households. Using changes in real incomes to update the 1984 half-median poverty line for other years, FaCS (2003, pp. 79–80) concluded that poverty had declined from about 11 per cent in 1984 to just over 5 per cent in 1998–99. Performing the same exercise with the 1984 half-mean measure also showed a decline in poverty from over 17 per cent in 1984 to less than 10 per cent in 1998–99.
Disposable income, discretionary income and housing
Almost all studies of income poverty use disposable income, which is the income after adjusting for taxes and government transfers. Government transfers include pensions, unemployment, disability and family benefits. Moller et al. (2003) reported a substantial reduction in poverty estimates in Australia after deducting taxes and government transfers. Poverty, defined by half-median income, declined from 16.2 per cent before taxes and transfers to 9.2 per cent post-tax, post-transfer.
Discretionary income and estimates of the poverty
Before and after-housing measures produce notably different estimates of poverty levels. According to the half-median income before-housing measure, with the modified OECD equivalence scale, 10 per cent of the population is in poverty. After deducting housing costs, the poverty rate increases to 15.5 per cent. Higher estimates of poverty, after taking into account housing costs, are also apparent on the other relative measures. On the half-median measure using the international equivalence scale, 12 per cent were in poverty before deducting housing costs, compared to nearly 16 per cent after deducting housing costs (Harding, Lloyd & Greenwell 2001, pp. 35–6). In contrast, the HPL does not show higher levels of poverty on the after-housing measure. Saunders (1996) reports data from the Australian Institute of Health and Welfare that shows a slightly higher level of poverty on the before-housing measure at 13.0 per cent compared to 13.8 per cent for the after-housing measure. Chotikapanich et al. (2003) reports substantially lower levels of poverty for the after-housing HPL measure compared to the before-housing measure.13 The reason for this discrepancy is that with relative measures, the poverty line is recalibrated with changes in the distribution of disposable incomes, whereas the HPL is based on fixed before and after-housing poverty lines that may not accurately reflect housing costs.
In contrast to the results for housing costs, estimates of poverty decline after non-cash benefits in health and education are taken into account. According to Smeeding et al. (1993), the estimated level of poverty in Australia (in 1981–82), defined by half the median disposable income, was only 7.4 per cent after taking into account these non-cash benefits compared to 15.1 per cent before adjustment.
Equivalence scales
Measures of poverty must adjust for household size. An income of $50,000 provides a very different standard of living for a single person than for a family of five. There are two common procedures to adjust for household size: the calculation of a multitude of poverty lines for different family arrangements and equivalence scales. One example of the first approach is the HPL, where the required expenditure for different family types was calculated using survey data on expenditure patterns in New York households in the mid-1950s (Johnson 1987).14 The Melbourne Institute of Applied Economic and Social Research publishes regular updates of the HPL. Separate poverty lines are calculated for ten different household arrangements with and without the household head working, and including and not including housing costs—giving a total of 40 poverty lines (for example, Melbourne Institute of Applied Economic and Social Research 2002). The ten household types do not cover all household types. The HPL is often referred to as the Henderson equivalence scale, although it is really a series of poverty lines for different family types.
More commonly, adjustments are made with equivalence scales that assign weights to additional adults and children. The modified OECD equivalence scale is becoming the most widely used equivalence scale. It assigns a weight of 1.0 to the first adult and a weight of 0.5 to the second and each subsequent adult. A weight of 0.3 is assigned to each child under 14 years of age (Förster 2001; Whelan, Layte & Maître 2002).15 The main disadvantage of this equivalence scale is that it does not incorporate economies of scale for a larger number of children or adults. Another popular equivalence scale that does incorporate economies of scale is the international scale, which simply takes the square root of household size (Organisation for Economic Co-operation and Development 1998; Osberg & Xu 2000, p. 6; Oxley et al. 1997).16 In Australia, the choice of equivalence scale changes the estimate of the proportion in poverty (see below). An international study based on data from the Luxembourg equivalence scale concluded that the choice of equivalence scale does not substantially change the proportion in poverty but does change the composition (Buhmann et al. 1988).
Equivalence scales and estimates of income poverty in Australia
Estimates of the level of poverty in Australia are sensitive to the equivalence scale employed. On the half-median disposable income measure before deducting housing costs, the proportion of persons living in poverty was 8.7 per cent with the Henderson equivalence scale, 10.1 with the modified OECD equivalence scale and 11.9 per cent with the international equivalence scale (Harding, Lloyd & Greenwell 2001, p. 35). In an international study of equivalence scales, Förster (1994) found that the estimates of poverty in Australia were particularly sensitive to the weight assigned to additional household members.
Zero and very low incomes
In most surveys of income, there is a small proportion of respondents who claim to have very low, zero or negative incomes. Logically, it is not possible for households to have little or no income for substantial periods. The reason for these responses may be that not all income sources have been documented or that there is considerable underreporting. Low-income households tend to underestimate government transfer payments (Oxley et al. 1997, p. 59). FaCS (2003, p. 91) reported that the aggregate level of income support benefits in the SIHC was about 20 per cent less than expected.17 Excluding households with unrealistically low or zero incomes would reduce the estimate of the proportion in poverty.