Endnotes
1. For example, Sen stated 'Indeed it could be argued that only in the absence of direct information regarding the satisfaction of the specific needs can there be a case for bringing in the intermediary of income, so that the income method is at most a second best'. (Sen 1982 in Mack, J & Lansley, S, Poor Britain, 1985). Ringen also stated that 'The relative deprivation concept of poverty is a direct concept: poverty is understood as visible poverty, that is, a low standard of consumption'. (Ringen 1988).
2. Internationally this finding has been reflected in the criticisms of Ringen and others as cited above. Australian analysis includes that reported in Hardship in Australia (Bray 2001).
3. In taking this stance FaCS notes that such improvements in living standards need to take account of the environmental and social outcomes in addition to achieving economic growth. Components of this of particular relevance to FaCS are the need for families to achieve an appropriate balance between their work and family responsibilities and the contribution that stronger communities make to household living standards.
At the same time, reflecting the policies of successive governments, and overall community aspirations, FaCS rejects arguments that economic growth and the resulting improvements in living standards reflect a consumerist treadmill that does not improve overall welfare.
4. Ian Castles, an Australian statistician and economist, has published various critiques of the United Nations Human Development Reports and the Human Development Index. Castles' critiques ranged across many aspects including: the need for accuracy, consistency and sound methodology in the selection and use of statistics for assessing progress and poverty reduction over time within particular nations (including the 'marginalisation' of particular countries); the comparison of levels and direction of poverty and human development between nations, the comparative rates (collectively) of social and economic progress of developed and developing nations; and the limitations and dangers of using single composite indexes—specifically the Human Development Index—rather than a range of relevant indicators for assessing human and societal development. The Human Development Index was a 'summary' index derived from three components: indicators of human longevity, education retention rates, and income per head (itself related to GDP). Underlying its development was a concern that 'narrowly economic' indicators such as Gross National Product or GDP alone failed to capture (qualitatively) important social dimensions.
5. Assuming earnings are spread equally over the year.
6. This type of confusion is unfortunately very common in analysis. For example Borland, Gregory and Sheehan (2002) cite the number of households of working-age in receipt of social security payments as 'a clear indicator both of broadening social problems and the impact of economic events on the social fabric' without reference to the extent this may have been driven by family payments. Similar analysis has been done by Birrell, Maher and Rapson.
7. This is sourced from FaCS modelling, and assumes that the two children are aged under five years, the family is not renting privately, and the income earner receives Male Total Average Weekly Earnings.
8. Youth jobless rate refers to persons aged 15-24. These figures are only available in the original series, and so February 1996 is compared to February 2003 to avoid problems with seasonality.
10. Male full-time ordinary time average weekly earnings.
11. These figures are higher than those used in the IGR as they include a number of smaller payments which were not incorporated into the IGR projections.
12. Family payments refer to the range of assistance provided to help with the additional costs faced by families with children. It has operated under several different names since its inception and is currently known as the Family Tax Benefit. The payment has a generous income and asset test, such that for the purposes of the quoted research, all but 15 per cent of the relevant birth cohort were eligible for inclusion in the study.
13. Teenagers include those aged 15 to 19 years.
14. In July 2002 20.9 per cent of 15 to 19 year olds were in the full-time labour market (original series); ABS 6203.0 July 2002.
15. An analysis comparing lone mothers receiving income support with those not receiving income support appears to indicate that lone mothers receiving income support generally have poorer outcomes than those not receiving income support across the range of barriers included in the table. However, lone mothers overall, whether or not receiving income support, appear to have poorer outcomes than partnered mothers for many barriers, and in particular for traumatic life events. Over 40 per cent of both income support and non-income support lone mothers experienced a 'traumatic life event', compared to less than 25 per cent for both groups of partnered mothers. This additional analysis has not been included in the text as the sample of non-income support lone mothers was quite small.
16. As at June 2002, there were 523,872 people receiving Newstart Allowance and 87,304 people receiving Youth Allowance who were not classified as full-time students, according to the FaCS Annual Report 2001-02.
17. Figures based on 2001 Census data. 'Not stated' responses were excluded prior to the calculation of percentages. Unless otherwise stated, proportions are of the relevant population aged 15 and over. In calculations of employment data, participation in CDEP has been treated as 'employment'. 'Occupation' data refer to the proportion of all employed people, aged 15 and over, in the relevant population, employed in specific occupations. Populations for tenure type (renting and home ownership) are based on households, not individuals. The 'families with children' data show the number of sole parent families with dependent children in the relevant population as a proportion of all families with dependent children in the relevant population. Dependent children include children aged under 15 and dependent students aged 15 to 24.
18. 'Multiple hardship' has been outlined earlier, and is discussed in depth in Bray (2001). The results for migrants are shown on p. 88 of that reference.
19. ABS 6203.0, Labour Force, December 2002, p. 43. Note that the results for Germany, Greece, Italy and The Netherlands have a relative standard error of 25 per cent to 50 per cent.
20. This includes the Age Pension, service pensions and other income support supplements.
21. The effective retirement age '… has been calculated as a weighted average of the various retirement ages where the weights are the probability of (net) withdrawal from the labour force at these particular ages'. (OECD, Employment Outlook, No. 72, 2002)
22. Centrelink Administrative Data, June 2002
23. FaCS estimates from Longitudinal Data Set.
24. All rates are current for the first quarter of 2003.
25. Tax returns are however only lodged by some 60 per cent of payers. It is therefore likely that some non- lodgers have low incomes, and that in many cases this is as a result of their receipt of income support with no supplementary sources of income.
26. Figure 15 shows unemployment as measured by duration since last full-time job.
In April 2001 the definition of duration of unemployment changed to refer to 'the period of time since a person last worked in any job for two weeks or more, regardless of whether it was full-time or part-time.' (ABS Cat. No. 6295.0) The questionnaire still collects data on duration of unemployment since last full-time job, allowing the time series data in Table 21 to maintain some continuity, however, other changes, including a redefinition from unemployed to employed of certain groups of people on unpaid leave, mean that data collected from April 2001 onwards are not strictly comparable with data collected in earlier periods.
27. The long-term unemployment figures differ from the numbers of people receiving unemployment payments for long periods. Nevertheless, the present favourable economic conditions and lower unemployment rates compared to the early 1990s are accompanied by a decline in the number of unemployed people on income support for a year or more. Between May 1999 and May 2002, the number of these customers dropped from 426,332 to 388,091. However, the number of unemployed people with income support durations of five years or more increased from 75,817 in May 1999 to 107,461 in May 2002. These figures are very difficult to interpret; unemployed income support recipients can have periods of full-time employment with no income support for periods of up to 25 weeks and still be regarded as long-term unemployed for the purposes of the administrative system, as documented in OECD 2001, Innovations in Labour Market Policy: the Australian Way, p. 185.
28. Figures obtained from ABS (6224.0) Labour Force Status and Other Characteristics of Families (1980-2000) and Labour Force 6291.0.40.001 companion data cubes (2001-02). The indicator 'children without an employed parent' presented under the economic disadvantage dimension shows substantial volatility over time that is not significant. The series is derived as a by-product from processing the ABS Labour Force Survey. However, the weighting applied in the production of the labour force estimates does not benchmark to the number of children in households in Australia (it is not required for labour force estimation purposes). One result is that much of the variation from year to year in the by-product series generated for children in households without an employed parent simply reflects variation in the coverage of all children in Australia.
29. FaCS modelling using Longitudinal Data Set 1 per cent sample 2001.
30. March to March comparisons are used for reporting changes in the labour market experience of youth as the data are highly seasonal and ABS do not publish seasonally adjusted estimates.
31. Refer also to discussion on page 98 concerning some of the limitations of these comparisons given the extent of variation in employer contributions for social insurance.
32. Data equivalised using revised OECD equivalence scale. To enhance comparability over time negative income from own business and property investment has been set to zero at the person level for data since 1994-95. In addition where an imputed tax liability exceeds household income the resultant income has been treated as a zero and not a negative amount.
33. Data comparability over this period is affected by three main factors:
a) The basis of 'benchmarking' of the survey has been changed. Some analysts such as NATSEM have considered this to be a severe enough issue for them to re-estimate the survey weights for their analysis.
b) ABS have changed the way in which they have treated negative incomes; while these were collected in 1990 such negative components were not used in the calculation of household income; in the later surveys the data were collected and used in estimating incomes. While conceivably this difference in approach can be overcome in analysis by applying inclusion or exclusion across the whole time period (as has been done in Table 7) this is not a comprehensive response to the problem as it appears that the 1990 data have some marked outliers in the distribution of negative incomes. It is likely that this has arisen as a result of less rigorous reviewing by ABS of these results in the earlier survey where the data were not actually used to determine the actual estimates of income. The impact of this can be seen in comparing three estimates of the changes between 1990 and 1994-95, firstly using the methodology as adopted in each of the years and the basis on which data have been published, the second based upon the 1994-95 methodology, and the third using the 1990 approach. For 'all households' these three approaches give estimates of income growth of 3.7 per cent, 4.9 per cent and 4.2 per cent; for lowest quintile households the rates are: 7.4 per cent, 8.3 per cent and 4.2 per cent.
c) ABS changed their survey methodology, including no longer asking people to extract pay and other records prior to interview, relying rather on people's ability to remember income details.
34. Statistical Local Areas are a geographic classification used by the ABS. They are either complete or partial local government areas.
35. It is important to distinguish between relative income poverty and income inequality. The former can be thought of as being concerned with the position of those on low incomes vis-à-vis the bulk of society, while the latter is concerned with the relative income levels of all groups in the society. Thus an increase in the incomes of the top 1 per cent would generally be thought to increase inequality without significantly increasing relative poverty, although the half mean approach would suggest that it did.
36. The US currently uses two different poverty lines: one for statistical purposes and the other for program purposes. The lines are however quite similar. The Census (or Orshansky) line is based upon three times the cost of a minimum food budget.
37. The initial research undertaken by Townsend into deprivation measures was largely directed at trying to derive such a point empirically. Townsend hypothesised that while it would be expected that deprivation had an inverse relationship with income, at some point the relative incidence of deprivation would increase rapidly as income fell and that this would establish a poverty line. It is generally considered that he, and subsequent researchers, have been unsuccessful in identifying such a point.
38. Disposable income, OECD revised equivalence scale and negative incomes set to zero at the person level.
39. Disposable income, OECD revised equivalence scale, negative incomes set to zero at the person level and imputed tax greater than income set to equal income.
40. While some of the items, such as increased housing wealth, can lead to higher levels of household consumption, taking this into account can only be done in much more sophisticated modelling which allocates estimates of flows of benefit from imputed rent, etc. This type of analysis is rarely ever done using the HPL.
41. In statistical terms the shape of the income distribution is usually considered to approximate a log-normal distribution.
42. The Sen index is the average of the headcount and poverty gap measures, weighted by a Gini coefficient that treats the poor as the whole population. Formally, it is expressed as:
where P0 is the headcount ratio, P1 is the poverty gap, and g is the Gini coefficient. Foster, Greer and Thorbecke's measure is a generalisation of the poverty gap measure. Formally, it is expressed as:
where x is the measure of living standards, z is the poverty line and 〈 is some positive parameter, typically α=2 (Deaton 1997, Chapter 3).
43. Given this, and other factors, it is often argued that zero and negative incomes should be deleted from such analysis, for example DeFina and Thanawalla, 2002. While offering a partial solution such approaches are essentially arbitrary as the question of data validity is as much an issue for many other households with very low, but positive, incomes. That is, the arguments for deleting a household say with $1 a week income would seem to be as valid as excluding one with zero weekly income.
44. Lancaster and Ray looked at various types of income pooling, and tested them on Australian household income and expenditure data. The results supported income pooling for some items but not for others, and found that income pooling across gender appeared much less likely for old people than younger ones. Lundberg, Pollak and Wales carried out a natural experiment analysis of a United Kingdom policy change, which transferred a substantial child allowance to wives in the late 1970s. They found strong evidence that a shift toward greater expenditures on women's and children's clothing relative to men's clothing coincided with this income redistribution.
45. This type of result is also reported by Ann Harding (1993).
46. The main exception to this near consensus comes from analysts using a rational choice framework. In this framework it can be argued that the decision of a couple to have a child is based on a belief that this will increase their overall level of welfare. That is, the couple decides that the benefits of having a child, including the happiness this provides them, outweigh the cost of the child. Similarly if a couple decide to separate, this is based on decision that they would be better off living by themselves. The use of equivalence scales in both these cases would suggest a reduction in wellbeing or welfare. Using the OECD scale for example having a child would reduce the apparent welfare of the household by 17 per cent = (1+0.5)/(1+0.5+0.3)), while, assuming the separation had no impact on incomes of either of the members, both of whom earn the same amount, the use of equivalence scales would suggest that they are both only 75 per cent as well off as they were when they were living together.
In part this type of example reflects the significant limitations in using income as a measure of wellbeing, and the importance of understanding the counter-factual case. In addition it needs to be noted that the judgements in the case of the couple having a child simply reflect the change in wellbeing of the couple, and not the child itself.
47. There is considerable variation in age cut off for a child in different analyses.
48. The main simplification of the scale was undertaken by Henderson himself; the major changes were to remove the gender-specific rates for children and the age-variant rates for both children and adults (Henderson 1975, pp. 355-6).
49. Refer Bray 2001; more detail on the results of analysis of this survey is presented in the discussion on deprivation and hardship.
50. For this table: Income support figures from Centrelink's A Guide to Commonwealth Government Payments, 20 March-30 June 2002; HPL estimates for single and couple adults (with exception of 'sharers' and Austudy recipients) from Melbourne Institute of Applied Economic and Social Research's Poverty Lines: Australia, June Quarter 2002, all other HPL estimates derived by FaCS; 50 per cent mean and median estimates based upon 1997-98 income data (ABS 2000c), adjusted for CPI changes.
Income support figures include the following assumptions: Pharmaceutical Allowance has not been included; individuals are receiving maximum rate of payment; students are receiving the 'independent' rate of Youth Allowance; the first three children in any family are under 13 years of age; the fourth child in a family is aged between 13 and 15; the youngest child is under five for families of three and four children and over five for families of one and two children.
51. In June 2002, only 23.3 per cent of income units receiving qualifying income support payments received Rent Assistance. (Data Source: Housing File, FaCS, June 2002)
52. It is generally considered that the degree of under-claiming for income support payments is low. It is very difficult to quantify take-up, even for programs where the major eligibility criteria are family composition, gender or age, and much more so where eligibility is determined in more complex ways, such as DSP or Newstart Allowance. A major reason people fail to take up an entitlement is because their payment level would be very low, and they do not consider it worth the effort. This could be true of significant numbers of people without being significant in monetary terms.
53. The major exceptions are those that quote child poverty rates only, which often use the Innocenti report as their source. In this study the researcher used his own estimates, directly derived from slightly later 1996-97 ABS data. It is not clear to what extent this data had been revised to maintain consistency with the edits undertaken by LIS to other countries. One important difference was that for Australia weekly data were used, whereas other nations were assessed on the basis of annual data.
54. These are the results currently posted on the LIS internet site as key figures (<http://www.lisproject.org/ keyfigures>). Two series of data are available, poverty estimates for adults, children and persons using poverty lines from 40 per cent to 60 per cent of median income, and inequality estimates using the Gini coefficient, two Atkinson estimates and three percentile ratios. Estimates are provided for each of the surveys that LIS holds.
55. The compulsory employer superannuation contribution since 1992 is of course equivalent to a social insurance tax, albeit at a much lower level than in most European countries.
56. One claim, which has had some currency in arguments that lower levels of inequality should be sought, was an apparent association between the rate of inequality and health outcomes. These were reviewed in the January 2002 British Medical Journal, along with a review article which concluded, 'It has recently become clear that the findings in [Wilkinson's] paper were an artefact of the selection of countries. Now that good data on income inequality have become available for 16 western industrialised countries, the association between income inequality and life expectancy has disappeared … the evidence for a correlation between income inequality and the health of the population is slowly dissipating'. (Mackenbach 2002)
57. This approach reflects the welfare economics approach of declining marginal utility of money. That is, an additional dollar will produce higher welfare for a poor person rather than a rich person. From this it can be argued that, under equal individual utility curves, income equality would maximise welfare. This though assumes a fixed amount of income across the society. Since it would involve the heavy transfer of income from those with the capacity to gain higher incomes, these would have no incentive to earn above the average, as they would gain little or nothing for their effort. This would in turn lower the total income of the society—and hence risk a loss in welfare for everyone. Hence the concept of 'optimal' tax theory—how far can redistributive taxes go in improving welfare before they act as a disincentive and generate a loss to total welfare.
58. The General entropy measure with α = 2 becomes half the squared coefficient of variation.
59. In such cases the two Lorenz curves will cross over each other and what is called 'Lorenz dominance' is not obtained, so it is not possible to draw any conclusion about the relative state of inequality.