Researching poverty: Methods, results and impact

Abstract

This article draws on evidence generated in recent deprivation studies conducted by the author and colleagues at the Social Policy Research Centre. After outlining some of the main limitations of poverty line studies, the paper explains how the deprivation approach addresses these weaknesses and illustrates the insights that deprivation studies can provide into the nature of poverty in contemporary Australia. It then compares the results produced by a deprivation approach with those produced using a poverty line – both in terms of what they imply about the extent of the problem and who they suggest is most affected by it. The comparisons demonstrate that the reservations that many hold about poverty research can be overcome and that when this is done, the results become more compelling and thus have the potential to have a greater impact on anti-poverty policy.

Introduction

Although poverty has not featured prominently on the Australian policy debate since the late-1980s, when Bob Hawke promised to ‘end child poverty by 1990’, it remains a topic of intense public interest. This is evidenced by the support expressed at the 2020 Summit for a national poverty reduction strategy (Department of Prime Minister and Cabinet, 2008) and by media and public reaction to a report on poverty released recently by the Australian Council of Social Service (ACOSS) to mark Anti-Poverty Week (ACOSS, 2012). The ACOSS report used an ‘international poverty line’ set at 50% of median income to show that despite strong growth in real average incomes, close to 2.3 million individuals, including 575,000 children were living below the poverty line in 2009–2010, after allowing for housing costs.

The report also showed that the overall poverty rate varied between 12% and 14% between 2003–2004 and 2009–2010 but showed no clear trend. The finding that poverty had hardly changed over a period when real average incomes had grown considerably attracted great media interest. Its findings were condemned by some as ‘a national disgrace’ and attacked by others for misinterpreting the meaning of ‘real’ poverty and exaggerating its dimensions. These debates are not new nor are the specific strengths or limitations raised by the supporters and critics of the methods used to measure poverty. At one level, the ongoing controversy illustrates the deep-seated feelings that the word poverty unleashes. At another, it highlights the challenges facing poverty researchers, who must confront these issues and resolve the inherent contradictions that they imply.

There have been a number of important developments in the international literature on poverty measurement over the last two decades that have implications for debate in Australia about the extent of poverty, its measurement, causes and consequences. Poverty line studies like those used in the ACOSS report remain important but are acknowledged to require additional evidence to sustain the claim that poverty exists among those with incomes below the poverty line.

The deprivation approach to poverty measurement that builds on the pioneering study conducted by Townsend (1979) over three decades ago is now widely used by researchers to supplement poverty line studies, particularly in European Union (EU) countries (Ward, 2009; Whelan et al., 2003). Evidence on deprivation has been incorporated into the poverty targets set by the UK and other EU governments to guide their anti-poverty strategies. Despite its growing importance in Europe, the deprivation approach has received relatively little attention in Australia – a feature that is all the more surprising given the widespread concern that has been expressed about the problems with poverty line studies.

This article draws on evidence generated in recent deprivation studies conducted by the author and colleagues at the Social Policy Research Centre (SPRC) (Saunders et al., 2007; Saunders and Wong, 2012). One of its goals is to illustrate the insights that such research can provide into the nature of poverty in contemporary Australia and to show how the results differ from those produced using a poverty line approach. Another is to demonstrate that the reservations that many hold about poverty research can be overcome and that when this is done, the results become more compelling and thus have the potential to have a greater impact on anti-poverty policy.

The following section provides a brief review of the recent literature on poverty, focusing on the differences between poverty line and deprivation studies. This is followed by a discussion of the data sources and methods used to estimate poverty and deprivation in Australia over recent years. The main results are then presented, highlighting differences in the estimates produced by the two approaches and exploring how they can be combined. The main conclusions are briefly summarised in the final section.

The parallel universes of poverty research

The Irish Combat Poverty Agency has proposed the following definition of poverty:

People are living in poverty if their income and resources (material, cultural and social) are so inadequate as to preclude them from having a standard of living which is regarded as acceptable by Irish society generally. (Combat Poverty Agency, 2004: 1)

The two key words in this definition are inadequate and acceptable. The first refers to the adequacy of the resources available and the second to the acceptability of the standard of living actually achieved. Although both components of the definition are important, they have generated two different strands of poverty research, as shown in Figure 1. The left-hand strand includes poverty line studies that examine the adequacy of economic resources in the form of income by comparing them with a benchmark poverty line. The right-hand strand, in contrast, examines the acceptability of living standards using as a benchmark the outcomes achieved, not the resources available.

Figure 1. The two strands of poverty research.

The deprivation approach fits within the latter strand of research because its focus is on establishing whether living standards are consistent with prevailing social norms and customary patterns of material consumption and behaviour. Its attractiveness rests on its more direct approach, avoiding the criticism labelled at poverty line studies that the poverty they measure is presumed to be an indirect consequence of low income, but this is never actually demonstrated. As Ringen (1988) has argued, the direct approach is more powerful because it avoids the assumptions (embodied in where the poverty line is set) that are implicit in the indirect approach.

Figure 1 has been deliberately drawn to show the virtual ‘parallel universes’ that poverty line and deprivation studies have occupied. In reality, this is a somewhat artificial distinction because many deprivation studies have been used to supplement the results produced by poverty line studies. Townsend himself argued that the income level below which deprivation rises steeply can be thought of as a poverty line. However, the method was widely criticised (e.g. by Piachaud, 1981), and few have subsequently followed this line of enquiry, reinforcing the gulf between the two approaches shown in Figure 1.

Recent developments in poverty research have begun to combine aspects of the poverty line and deprivation approaches, as shown at the bottom of Figure 1. This has taken the form of two developments. From within the poverty line tradition, growing concern over the need for a definition of economic resources that is broader than income, coupled with a need to better demonstrate that poverty actually exists, has led to what is referred to here as attempts to apply an ‘adjusted poverty line approach’, or to validate the existence of poverty. From within the deprivation literature, the concept of consistent poverty has emerged that combines low-income with evidence that deprivation exists as a way of demonstrating that the standard of living is unacceptable.

Two variants of the adjusted/validated approach have been developed in the Australian context. The first has involved utilising data on financial stress to establish whether its incidence among households with incomes below the poverty line is consistent with those households experiencing the kinds of difficulties that are associated with poverty (see Bray, 2001). Since financial stress variables were included in the Australian Bureau of Statistics (ABS) Household Expenditure Survey (HES), they have been included in other ABS surveys (e.g. the General Social Survey (GSS)) and have been used by the ABS to identify who is struggling to make ends meet (ABS, 2012). A number of the indicators (and some others) have also been included in the Household, Income and Labour Dynamics in Australia (HILDA) survey (see Wilkins and Warren, 2012: Chapter 9). Studies using these data have found that while those with incomes below the poverty line experience more instances of financial stress on average than those with incomes above the line financial stress also exists among those in the upper sections of the income distribution, raising concern about the ability of the financial stress indicators to identify poverty (as opposed to financial mis-management or poor budgeting).

The second development is more recent and has involved supplementing income with other measures of economic capacity or resources as a way of better measuring economic well-being generally and, from that, providing better estimates of poverty. Work using the framework developed by Headey (2006), for example, has been used by the ABS to produce a measure of economic resources that combines information on income and wealth to better capture the consumption possibilities available to households (ABS, 2009a). ¹ Another recent application of the ‘combined indicators’ approach is in the social inclusion indicator framework that has been developed by the federal government to monitor improvements in social inclusion of households facing multiple disadvantage – defined as those experiencing a combination of low income, low wealth and financial stress (Australian Government, 2012).

These developments address one of the weaknesses of poverty line studies (the reliance on income alone to measure economic capacity) but many other limitations remain. Among the most important of these is the critical issue of where to set the poverty line – a decision that will affect not only how many are poor but also the composition of households identified as poor. Another key decision revolves around the choice of equivalence scale. This is used to adjust household income for the needs of household members and has an important bearing on which kinds of households are identified as poor – in terms of their overall size and composition (e.g. adults vs children).

Most poverty studies now use the modified Organisation for Economic Cooperation and Development (OECD) scale, which assigns a value of one to the first adult in each household, 0.5 to each subsequent adult and 0.3 to each dependent child (assumed to be aged below 15 years; older children are treated as additional adults). ² The OECD and other widely used scales have an important bearing on estimated poverty rates among children and older people (who mainly live in adult-only households). The differences between poverty rates among children and older people have important policy implications for the relative generosity of family payments and pensions, but it is important to bear in mind that such comparisons are a direct consequence of the equivalence scale used to produce the poverty estimates.

One limitation of the modified OECD scale is that it assumes that the needs of all adults are the same, irrespective of their age, gender and other factors. Thus, a 20-year-old male is assumed to have the same needs as a 70-year-old male, and both are assumed to have the same needs as a 35-year-old female with a disability. The example illustrates the potential importance of the assumptions built into the equivalence scale. In addition, the fact that the OECD (and most other) scales are not based on empirical studies of household need introduces another arbitrary element into the results and provides another reason to ignore them – or at least to minimise their impact on policy deliberations.

The impact of these issues can be examined by conducting sensitivity analysis, in which the impact on the estimates of varying the poverty line (e.g. from 50% to 60% of median income) and varying the equivalence scale (e.g. by replacing the OECD scale by the square root scale) and such analysis is (or should be) an important component of any robust poverty line study. However, sensitivity results tend to detract attention away from the ‘headline results’, send a confusing message about key findings and reinforce the concern that the estimates do not provide a solid foundation on which to change policy.

However, despite all of their limitations – most of which were acknowledged long ago – poverty line studies remain important as a way of highlighting who faces the greatest risk of poverty. This in turn has implications for the adequacy of existing levels of the key variables that affect the discretionary incomes of those at greatest risk, including the levels of social security benefits and the minimum wage, and of unavoidable living cost such as rent, basic food items, out-of-pocket medical expenses and household utility bills.

In contrast to the poverty line approach, the starting point of the deprivation approach involves identifying items that are considered necessary or essential for everyone to have in order to achieve an acceptable standard of living. Details of how the approach is operationalised are provided in a recent article in this journal (Saunders and Wong, 2011: section 3) and are summarised only briefly here. The key point is that the items must receive majority support among members of the community for being essential, so that an inability to acquire these items is indicative of having an unacceptable standard of living, where community views determine the precise items that represent that acceptable minimum. Importantly, because the decision about which items are essential is taken by community members (as expressed in responses to a social survey) not by ‘experts’, this gives the approach a better grounding in everyday experience and the lessons learnt about what items are required to meet basic needs by those best judged to distil them.

Once the essential items have been identified, deprivation is defined to exist when people do not have these items because they cannot afford them. ³ It is possible to sum the numbers of items that each individual is deprived of, and the resulting index score is a simple but intuitively appealing measure of the severity of deprivation that can be averaged across different groups and compared. A higher index score indicates higher deprivation, but it is important to note that no threshold has been used to separate those who are deprived from those who are not ⁴ nor have any assumptions been made about the needs of each individual or household.

To return to the example cited earlier, if the needs of the 20-year-old and 70-year-old males and the 35-year-old female with a disability are the same and if they have access to the same level of resources, they should all face the same level of deprivation. If, in contrast, the level of deprivation faced by the three individuals differs, then it follows that they differ in their ability to meet their basic needs from the resources available to them. If their incomes are the same or very similar (e.g. if they all rely on the same level of social security payment), then it follows that their needs differ. These conclusions do not depend on assumptions made about relative need (using an equivalence scale) because under the deprivation approach, the relevant information can be inferred by examining how deprivation varies across different individuals, households, or social groups.

If it is thought useful to set a threshold that differentiates between those who are deprived and those who are not, this can take the form of a minimum number of essential items that are missing and cannot be afforded. The percentage of households that exceed this number then corresponds to the poverty rate, but it too is dependent on where the deprivation threshold is set, and this requires a judgement to be made – unlike the index score approach that avoids the need to make this judgement. ⁵

It should by now be apparent that the deprivation approach has many advantages over the conventional poverty line approach. It avoids the need to make a judgement about where to set a poverty line threshold or to make an assumption about how needs vary with household size and composition. By relying on community opinion to identify which items are essential, the deprivation approach is grounded in practical experience and reflects an everyday understanding of what poverty means: going without basic items. Finally, the deprivation approach adopts a definition of what is needed to achieve an acceptable minimum that is consistent with prevailing community norms and standards of acceptability.

However, it does not follow that the poverty line approach should be dispensed with altogether. Income remains an important determinant of the standard of living and most government programmes that seek to alleviate poverty do so by either providing income transfers directly to households, or by improving access to income streams (e.g. earnings, or interest income) that can be used to meet basic needs. It follows that there is value in the conventional poverty line approach, as long as its limitations are acknowledged and, where possible, addressed or subjected to sensitivity analysis. This has led to the development of the concept of consistent poverty (see Figure 1) that exists when income is below the poverty line, and a minimum number of essential items cannot be afforded.

The consistent poverty approach was first developed in Ireland, where it was adopted by the Irish Government as one way of assessing the success of its anti-poverty strategy (see Nolan, 2000). A similar approach was subsequently adopted in Britain to monitor the child poverty pledge introduced by Tony Blair, and the approach is now widely used throughout the EU as a way of documenting and monitoring poverty trends. The approach has been applied in Australia by Saunders and Naidoo (2009) and more recently has been used by Saunders and Wong (2013) to estimate the social impact of the global financial crisis.

The consistent poverty approach shares many features with the other approaches that have been applied in Australia described earlier. In particular, the importance of having access to a minimally adequate level of income is maintained, but other information (on financial stress or the experience of deprivation) is used in combination with information on income to better identify who is in poverty. The extensions to the conventional (income-only) approach reflect the limitations of relying solely on income, but they also have the effect of making the results more credible by showing that low income is accompanied by the kinds of adverse outcomes that are expected to be a consequence of living in poverty. To the extent that they succeed in this objective, the new methods produce estimates that are more likely to be taken seriously by policymakers and are thus more likely to have an impact on the policies introduced to address poverty.

Data and methods

Data from the ABS Survey of Income and Housing (SIH) conducted currently every 2 years are used in most academic studies of poverty (e.g. Phillips and Nepal, 2012; Saunders and Hill, 2008; Wilkins, 2008) and are also the source of the ABS’s own income distribution reports (e.g. ABS, 2011). Increasingly, poverty studies are utilising the HILDA data to study poverty either statically (e.g. Wilkins and Warren, 2012: chap. 7) or to examine the dynamics of poverty (e.g. Buddelmeyer and Verick, 2008). Although the HILDA data have much to contribute, the ABS data remain the most authoritative source of cross-sectional household-level income data, although a number of recent changes to survey methodology, to data manipulation (e.g. benchmarking techniques and weighting) and – most significantly in the current context – in the definition of income have undermined the ability to use the ABS data to track changes over time in poverty (and inequality more generally).

Among the most important of the recent changes was the inclusion of salary sacrificing in wage and salary income (introduced in 2005–2006) and of irregular payments such as bonuses, redundancy and irregular overtime, introduced in 2007–2008. Analysis conducted by the ABS indicates that the latter change affected 3.4 million (43%) households, resulted in an increase in mean income by $85 a week (5%) and caused the Gini coefficient to increase from 0.317 to 0.331, or by 4.4% (ABS, 2009b; see also Kindermann and McColl, 2012). The ABS provides information on how these changes affect income (by providing income estimates both before and after the change), but this is limited to only a few years, making it impossible to produce a consistent time series that spans longer periods. For current purposes, the estimates presented later have been derived using the income definition that prevailed in 2005–2006, since this is available in the 2 years that are the focus of this analysis (2005–2006 and 2009–2010).

Poverty has been estimated by comparing weekly disposable (equivalised) income with a poverty line set at 50% of (modified OECD) equivalised disposable income. ⁶ The estimates have been person-weighted so that they refer to the numbers of persons in households with incomes below the poverty line. ⁷ Poverty has been estimated both before and after housing costs – a distinction that was common in earlier Australian poverty studies, but one that has tended to be ignored in recent studies. ⁸ The after housing cost estimates involve comparing income after deducting housing costs with a revised poverty line set at 50% of the median value of the distribution of income minus housing costs. ⁹

The source for the deprivation estimates is the two SPRC surveys of poverty and social disadvantage conducted in 2006 and 2010. These have been described extensively elsewhere (Saunders et al., 2007; Saunders and Wong, 2012), and only those aspects that are relevant to this analysis are repeated here. Each survey asked participants whether a range of items – 63 in 2006 and 71 in 2010 – are essential ‘for all Australians’, where essential was defined as ‘things that no-one in Australia should have to go without today’. A total of 24 items received majority support for being essential in both surveys, and these have been used to identify deprivation, which exists when people do not have these items because they cannot afford them. ¹⁰

The 24 items can be categorised into the six broad areas of need shown in Table 1. ¹¹ A deprivation score was derived for each responding household by summing the number of deprivations, and these can be averaged and compared across household types in order to establish which groups are most deprived. The resulting comparisons correspond to the familiar poverty rates by household type, and the focus of the results presented in the following section is on how closely the two sets of estimates align and which groups are found to be most at risk according to the two approaches.

Table 1. Need classification of the essentials of life in 2010.

Before turning to those comparisons, it is worth observing that 5 of the 24 items shown in Table 1 relate specifically to the needs of children. Although these are important aspects of basic needs, their inclusion in the analysis means that households that contain children (defined in the deprivation analysis as those aged below 18 years) can potentially attain a higher maximum deprivation score than those without children, thus distorting to some degree the comparisons between those with and without children. In order to allow for this problem, the five child-related items have been excluded and the deprivation scores re-calculated.

Results

The results produced by the methods described above are presented in Table 2, which shows the patterns of poverty and deprivation by household type. ¹² The before housing cost poverty rates in both years for those aged 65 years and above are very high, particularly for single older people, who face the highest overall poverty rate: almost four times the national rate in 2005–2006 and over three times the overall rate in 2009–2010. However, these patterns change markedly when account is taken of housing costs. On an after housing costs basis, as the rankings (shown in brackets in Table 2) indicate, poverty among older people is much lower, and the issue of child poverty now becomes more apparent.

Table 2. Estimates of poverty and deprivation by household type, 2006 and 2010 (household rankings shown in parenthesis where 1 = highest rate). ^a

WA: working-age; OP: older person.

a Poverty rates are for 2005–2006 and 2009–2010, deprivation scores are for 2006 and 2010.

b WA refers to those aged 15–64 for the poverty rates and to those aged 18–64 for the deprivation scores.

c OP refers to those aged 65 and above in both cases.

Although there is little change over the period in the overall poverty rate on both a before and after housing costs basis, this is not the case for all household types, with working-age couples (with or without children) faring worse in 2009–2010 than in 2005–2006, particularly once account is taken of housing costs. The other notable feature of the poverty estimates in Table 2 is the marked improvement in the poverty status of older people – particularly single older people – between 2005–2006 and 2009–2010. This is a consequence of the substantial increase in the single rate of pension that followed the recommendations of the Harmer Pension Review (Harmer, 2009) that began to be paid in September 2009. ¹³

In contrast with the poverty rates, the household rankings in Table 2 show that the deprivation scores are highest for sole parent families, followed by single working-age people in both years. This is the case, irrespective of whether or not the five child-related items are included. ¹⁴ Older Australians (particularly older couples) now show up as being less deprived than other groups and as facing a level of deprivation that is below the overall rate. Deprivation is lower among single older people in 2010 than in 2006 whichever measure is used, and this again points to the impact of the pension increase. It is also important to note that the deprivation results show a clear decline over the periods, unlike the poverty rates that are broadly unchanged.

In overall terms, the deprivation results thus not only show greater improvement over the period but also suggest that the problem is most prevalent among families with children and single people of working-age. In contrast, the poverty rates show no sign of improvement over the period but also indicate that the problem is greatest among older people. These contrasting findings have important implications for policy. In particular, they suggest that the deprivation approach is better able to pick up changes over time because, unlike poverty rates, they do not automatically shift upwards whenever real incomes (and hence the poverty line) are rising.

This does not mean that deprivation is independent of community living standards (which it clearly is not), but it does highlight the fact that deprivation is more firmly rooted in the everyday realities of those who are doing it tough. Rising incomes of the middle classes and the rich will eventually cause the items considered essential to change (and possibly also their cost and affordability), but these factors are not so deeply embedded in the measurement process as they are when poverty is measured using a relative poverty line anchored to median income. ¹⁵

This factor, along with the other advantages of the deprivation approach discussed earlier, suggests that the deprivation results in Table 2 need to be taken seriously, and the problems they point to need to be addressed in any overall poverty reduction strategy. The estimates in Table 2 highlight the limitations of poverty line studies, particularly in a period of rising real incomes, but they also illustrate how a poverty line approach can produce an overall poverty rate that is difficult to shift. This can also affect the composition of the poverty population in ways that can send the wrong messages to those responsible for policy.

One important lesson to emerge from Table 2 is that measuring poverty on an after housing costs basis produces results that are closer to those produced by the deprivation approach. To the extent that the deprivation estimates are better able to identify instances where resources available are not able to meet existing needs (as this article argues), it follows that account should be taken of housing costs when estimating poverty using a poverty line approach.

Conclusion

The primary focus of this article has been methodological. It has discussed two approaches to poverty measurement: one based on the use of a poverty line and the other on asking directly whether people can afford items that are widely regarded as essential. The poverty line approach has many limitations, including that the existence of a low income is presumed to result in poverty. However, that presumption is not itself subject to further examination (as it is when poverty estimates are combined with evidence on financial stress). For this reason, poverty line studies are best thought of as estimating the risk of poverty. Further problems have arisen in the Australian context as a result of improvements in data quality that have undermined the ability to examine how poverty has changed over time.

The deprivation approach is linked more directly to living standards and seeks to establish whether people can afford the items that are widely regarded as essential in today’s society. Those items are identified on the basis of community views, and this takes a crucial element in the measurement process out of the hands of experts and places it (rightly) in the hands of members of the community. The focus on a lack of affordability makes the deprivation approach aligned with poverty, where it is a lack of economic resources that ultimately determines whether someone is poor. However, this is established by assessing people’s ability to acquire essential items, not by comparing their income with a poverty line.

Both approaches have their merits, but the basic argument advanced in this article is that poverty line studies alone are not sufficiently credible to persuade policymakers that the problem identified is ‘real’ and in need of action. The deprivation approach – in isolation or in conjunction with a poverty line approach as reflected in the concept of consistent poverty – has the potential to generate robust and credible results that are far harder to discredit and thus more capable of having an impact on understanding of the issue and what is needed to address it. It is in this sense that deprivation studies have the potential – as yet largely untapped in the Australian context – to shape the debate over poverty and have a positive impact on the lives of those who are experiencing it.