From Multiple Deprivations to Exploitation: Politicizing the Multidimensional Poverty Index

Abstract

OPHI and UNDP, Global Multidimensional Poverty Index 2022: Unpacking Deprivation Bundles to Reduce Multidimensional Poverty. Oxford and New York: Oxford Poverty and Human Development Initiative and United Nations Development Programme, 2022. 39 pp. https://ophi.org.uk/global-mpi-report-2022/

Few questions are more fraught, or more consequential, than the number of people in the world who are poor, and whether that number is rising or falling. There is no shortage of high-profile liberal ideologues who, in recent years, have been happy to claim that global capitalism has managed to drive down the global poverty rate from some 90 per cent at the turn of the 19th century to around 10–15 per cent today. This claim was made perhaps most (in)famously in recent years by Steven Pinker in his 2018 bestselling book Enlightenment Now. Pinker's data and claims about poverty draw heavily on economist Martin Ravallion's work,1 although he notably brushes aside the latter's caveats and methodological caution in favour of burnishing a teleological narrative of inexorable rationalization, enlightenment and progress.

Such claims echo a longer history of optimistic readings of the capacity of neoliberal capitalism to counter poverty2 and are overwhelmingly based on income-threshold measures of extreme poverty that Ravallion and colleagues originally helped to develop and popularize in the late 1980s and early 1990s (Ravallion et al., 1991). This approach to measuring poverty has long been contested (see Wade, 2004). For instance, based on other measures of poverty such as estimates of real wages and data on height and mortality rates, Dylan Sullivan and Jason Hickel (2023) have recently argued that the rise of capitalism has in fact worsened extreme poverty globally.

Such divergent assessments of the level and trajectory of global poverty are possible in part because measuring poverty is slippery business. Analysts must deal with patchy data, alongside thorny methodological and measurement problems, which are ultimately grafted on top of foundationally contested normative and political questions about what exactly poverty entails (see Fischer, 2018). Whether income thresholds are an adequate or meaningful way of understanding and counting poverty, especially on a world–historical scale, is first and foremost a normative question, even though it often masquerades as a technical problem.

Poverty measures are no less contested at lower levels of aggregation, however. They are invariably both objects of political contention and tools of statecraft. Katharina Lenner, for instance, shows how the contested construction of poverty measures in Jordan simultaneously renders poverty intelligible to the state while also ‘obscur[ing] worsening socio-economic situations, and deflect[ing] responsibility for the situation away from government offices’ (Lenner, 2023: 2). There is no escaping the politics of poverty, and measuring poverty is not an innocent act. Neither, for that matter, are any other measuring acts (see Christophers, 2013; Hansen and Porter, 2012; Mügge, 2022).

Enter, against this backdrop, the subject of this Assessment: the Global Multidimensional Poverty Index 2022: Unpacking Deprivation Bundles to Reduce Multidimensional Poverty (hereafter MPIR 2022 or the Report), the latest report on the Multidimensional Poverty Index (MPI). Based on earlier work by Sabina Alkire and colleagues at the Oxford Poverty and Human Development Institute (OPHI),3 the MPI was launched in 2010 as an annex to the Human Development Report (HDR) on the Human Development Index (HDI), published annually by the United Nations Development Programme (UNDP). The MPI was significantly revised in 2014 and 2018 in an effort to bring it in line with the Sustainable Development Goals (SDGs) and since 2018 it has been published as an annual thematic report unto itself. The basic orienting idea behind the MPI is that poverty is multidimensional — that is, poverty encompasses multiple separate but linked deprivations — and, as such, needs to be measured in a way that accounts for a broad spectrum of deprivations that are not reducible to income. Correspondingly, the MPI is not based on measurements of income, but instead classifies poverty according to a combination of indicators, including a lack of sufficient health care, nutrition, education, as well as a lack of access to housing, water, sanitation and the like, the latter grouped together under the heading of ‘standard of living’. Deprivations in terms of education, housing or food tend to be correlated to income and to each other, of course, but far from perfectly, and in ways that income measures do not effectively capture.

The MPIR 2022 estimates a global poverty headcount based on a composite index of these multiple forms of deprivation, although the authors are often at pains to stress that it is a complement, rather than an alternative, to the World Bank's income-based poverty estimates. Besides updating the index and headcount, the MPIR 2022 makes two specific contributions: it provides an estimate of the impacts of the COVID-19 pandemic on ‘multidimensional’ global poverty and introduces an analysis of ‘deprivation bundles’ based on the MPI data. A ‘deprivation bundle’ refers to the ways in which different forms of deprivation tend to co-occur, so that analyses can highlight common pairs or triads of deprivations that individuals or households frequently experience together, either at a global scale or in particular settings. The MPI in general, and MPIR 2022 in particular, merit a closer look because in important ways the MPI is illustrative of the wider framing of poverty in global development debates and discourses. While it is a cutting-edge piece of work and among the most thoughtful recent efforts to measure and govern poverty, it nonetheless is part of a wider global politics of poverty.

Considering multiple and interrelated types of deprivation offers no liberation from the thorny politics of measuring poverty more generally. Of course, the architects of the MPI themselves do know this — to their credit, they are often explicit and reflexive about the inevitably normative nature of choices on how to ‘count’ poverty, although this has tended to remain confined to considering the ethics of what counts as ‘poverty’ (see Alkire and Kanagaratnam, 2021). It's on this terrain that I want to engage with the MPIR 2022 in this Assessment by asking what, normatively and above all politically, is accomplished by measuring multidimensional poverty.

The main observation I wish to make in this article is that poverty as conceived in the MPI, as well as in imitator indices like the World Bank's Multidimensional Poverty Measure (MPM), is multidimensional, but not relational. The index and its construction explore the intersections of different forms of poverty, but almost by definition they hide from view, to borrow Bernstein's useful phrase, ‘the mechanisms that generate both wealth and poverty as two sides of the same coin of (capitalist) development’ (Bernstein, 1992: 24; see also Selwyn, 2014). In this sense, the MPI and like measures inevitably operate as a kind of fetish. They show how multiple forms of deprivation interact and attach a sophisticated quantitative metric to their depth and severity. All the while, however, they (deliberately or otherwise) obscure the social relations and histories behind these variegated forms of poverty. By looking at the index, we learn in fine-grained detail how many people are poor, where and in what ways, but without a sense of how and why they became and remain impoverished and how such processes of impoverishment are integrally linked to the accumulation of wealth and capital elsewhere. For this, I argue in the latter parts of this article, we need to conceptualize poverty as an outcome of exploitation — the complex set of social relations through which value is appropriated from nature and labour and accumulated as wealth or capital.

This ultimately matters a great deal when it comes to what is surely the most important question: what is to be done? (With due apologies to Lenin….) The MPI doesn't come packaged with an explicit political theory, but of course it has one. And it's easy enough to see what it is if we read between the lines. The MPI presents itself as a granular mapping of the landscapes of poverty, expressly in the service of developing more finely tuned and especially better-targeted policy measures. As shown below, the MPI implicitly entrenches a preference for means-tested interventions targeting the ‘poorest’ over universal or redistributive measures. It is a version of things largely devoid of consideration of the social relations linking processes of impoverishment to the accumulation of capital, which might elicit more politically challenging questions than the Report is keen to engage. Whether intentionally or not, in this way, the MPIR 2022 depoliticizes and ultimately naturalizes poverty, hiving it off from the question of wealth and capital accumulation and presenting its resolution as an issue of gradual, targeted technical intervention. Ultimately, I argue, to get around some of the political and practical blind spots that this kind of analysis creates, we need an approach to measuring poverty that centres on relations of exploitation as a cause of poverty.

The remainder of this essay develops these arguments in three sections. The first section below describes the background to the MPI within a wider politics of poverty, particularly since the turn of the century, and then discusses the index itself. The second section discusses the MPIR 2022, showing how the concept of ‘deprivation bundles’ at its core reflects and amplifies the depoliticizing tendencies implicit in the MPI more generally and how it amplifies implicit tendencies towards targeted rather than universalist approaches to poverty. The final section argues that poverty must be conceptualized in a way that accounts for its emergence and maintenance through relations of exploitation. This would help to move beyond current symptom-focused approaches towards accounting for causes of poverty, and help to repoliticize poverty in the process.

The MPI in important ways is a product of a particular form of global poverty politics that emerged in the 1960s and is still present. The overview of the index that follows will necessarily be somewhat brief and general but is still worthwhile in helping the reader grasp the wider politics at play in the process of constructing the index. As scholars have recently argued, the neoliberal ‘counter-revolution’ emerged in no small part in and through the failures of the ‘assault-on-poverty’ agenda pursued by the World Bank and other international organizations in the 1970s (see Bernards, 2022; Mendes Pereira, 2020).

Thus, where poverty had previously mainly been considered under the rubric of ‘social welfare’ in the core economies of the global North and ‘development’ in the global South, a complex and cross-cutting ‘global consensus’ on poverty and poverty reduction measures swiftly emerged at the turn of the century. Mkandawire (2005) rightly notes a related shift in poverty reduction measures, again both in the global North and global South, from universalistic social programming toward targeted and means-tested benefits. This is perhaps epitomized in the above-mentioned consensus on poverty. For Peck (2011: 176), this consensus was accompanied by a dynamic of ‘fast-policy development’ — the increasingly rapid rollout, spread, failure and consequent adaptation of various targeted interventions aimed at mitigating poverty. In the 2000s, conditional cash transfers and microcredit schemes were probably the prime examples. By 2023, these have largely fallen by the wayside in favour of more ‘modern’ interventions including digitalized social transfers, biometric identification systems and national financial inclusion frameworks, alongside interventions directed at improving ‘self-sufficiency’ at the local level, such as pay-as-you-go (PAYGO) off-grid solar electricity projects, and others focusing on ‘resilience’, such as the renewed push for agricultural modernization under the guise of promoting ‘resilience’ to climate hazards.

This is the context in which, for instance, randomized control trials (RCTs) of anti-poverty interventions proliferated. By the late 2000s, the ‘randomistas’ had quite successfully established themselves as influential actors in global development politics, and RCTs were widely seen as the ‘gold standard’ for evaluating development interventions. The reason for the rise of RCTs in many ways mirrored those for the wider ‘fast-policy’ landscape around global poverty. Budgets in developing countries were heavily constrained, particularly in the aftermath of structural adjustments. Combined with constant restrictions on aid budgets in the global North, this pushed both a ‘projectization’ of development aid and a growing reliance on frequently cash-strapped NGOs to administer projects. This provided fertile ground for a methodological orientation that promised cost effectiveness (by ‘rigorously’ identifying ‘what works’) and conceived of ‘development’ action as so many targeted, dispersed, small-scale, testable interventions (see Bédécarratts et al., 2019; Donovan, 2018; Kvangraven, 2020; de Souza Leão and Eyal, 2020).

The RCT fad can be seen as forming part of a wider push towards quantifying and targeting poverty at aggregate national and global levels. A combination of expressly global techniques and targets for tracking and measuring poverty has played a critical role both in driving and in sustaining the global consensus around poverty. The Human Development Report, first published in 1990 to reflect on the Human Development Index, has arguably been at the forefront of this shift toward the quantification of poverty (UNDP, 1990). Following the HDI were a growing number of benchmarking and measurement exercises centring on global poverty. The ‘goal-setting’ exercises of the Millennium Development Goals (MDGs) launched in 2000 and the Sustainable Development Goals that succeeded them in 2015 are perhaps the most notable and visible examples. The MDGs and SDGs probably need little further elaboration here, but it is worth noting that the very process of framing and measuring ‘extreme poverty’ and inequality in and through the MDGs and SDGs has long been contested on several levels. Numerous scholars, including Fischer (2018), Fukuda-Parr (2019), Gabay and Ilcan (2017) and Weber (2015), have previously pointed out how these targets have depoliticized poverty interventions.

The MPI should be understood as emerging from and forming part of this landscape. The initial articulation of the MPI was heavily influenced by Amartya Sen's critique of income-based measures of poverty and, of course, the HDI. The original working paper outlining the method and rationale for the MPI cites Sen's evocative comment, forming part of a reflection on the first decade of HDRs, that ‘[h]uman lives are battered and diminished in all kinds of different ways, and the first task, seen in this perspective, is to acknowledge that deprivations of very different kinds have to be accommodated within a general overarching framework’ (Sen, 2000: 18, cited in Alkire and Santos, 2010: 6).

The MPI was accordingly designed to map out the overlaps between different forms of deprivation at the household level, aggregating from there to national and global headcounts. This was an effort that from the start was framed as part of the wider global anti-poverty movement, reflected at the time in the MDGs and in the design of targeted poverty measures. The working paper describes the unique value of the MPI as follows: ‘providing information on the joint distribution of deprivations related to the MDGs — which shows the intensity and the composition of several aspects of poverty at the same time — we have tried to explore how better measures could support efforts to accelerate the reduction of multidimensional poverty’ (Alkire and Santos, 2010: 6). It is perhaps worth emphasizing, however, that particularly in a revised version of that paper, later published in World Development, the authors were keen to insist that the MPI ultimately ‘has a similar spirit to that which once motivated the “dollar-a-day” measure’ in that it aims to assess the magnitude of poverty in the developing world, and that in doing so across countries ultimately relies on avowedly ‘rough but methodologically consistent’4 measures (Alkire and Santos, 2014: 252).

The MPI and associated thematic reports in many ways are remarkable achievements. They represent the amalgamation of a huge volume of work, mostly coordinated between the UNDP and the OPHI. The construction of the index entails pulling together and analysing a truly staggering volume of data from large-scale household surveys in 111 countries, including those from the US Agency for International Development's Demographic and Health Surveys programme, UNICEF's Multiple Indicator Cluster Surveys and a collection of national exercises. Nonetheless, it clearly continues within and reflects the wider political consensus on poverty of the last 20 years. Indeed, it is not coincidental that the World Bank itself now publishes its own MPM, self-avowedly ‘taking inspiration and guidance from’ the MPI (Diaz-Bonilla et al., 2023: 2).5

The MPI was subject to minor revisions in 2014 and 2018. In the current (2018) version, poverty is calculated on the basis of 10 indicators that are grouped into three categories. Table 1 provides a more detailed breakdown of the current indicators and their definitions. Each of the three measures (Health, Education, and Standard of Living) accounts for one-third of the total index. The measures ‘Health’ and ‘Education’ each contain two sub-indicators (each weighing one-sixth of the total score), while the ‘Standard of Living’ component contains six (each weighing one-eighteenth of the total score). According to these indicators, a person would be considered ‘multidimensionally poor’ if they lived in a household meeting 33.3 per cent or more of the criteria of the weighted index (Alkire et al., 2022).6 Aggregate measures of poverty headcount and severity are extrapolated from the survey to the national level and are then used to make global estimates.

In short, poverty is a more widespread and complex phenomenon if we understand it as encompassing multiple dimensions. In the MPI, however, poverty is nonetheless naturalized and ahistoricized, framed as a timeless scourge at which heroic development interventions are gradually chipping away. The MPI itself importantly does not actually allow us to determine how such ostensible poverty reduction measures are enacted. However, as Fischer (2018: 121–22) points out, some of these might simply be effects of the measures of poverty themselves. The MPI, for instance, measures ‘standard of living’ by regarding a quite narrow and largely urban-biased range of consumer goods and infrastructures. Fischer (ibid.) notes, ‘improved’ sanitation facilities are vitally important for public health in urban areas, but often relatively less so in rural areas with lower population densities. Equally, the list of ‘assets’ considered in the index mainly includes household appliances such as refrigerators and electronics, but not (with the one exception of animal carts) those linked to agrarian livelihoods, such as livestock, equipment or land tenure, which might be particularly salient in rural settings (ibid.). One implication of this is that the index probably tends to overcount rural poverty. As a result, processes of urbanization might well produce reductions in the MPI poverty headcount without necessarily producing meaningful improvements in livelihoods or well-being. The presumption that reductions in MPI-measured poverty are being achieved via targeted interventions is both unfounded and doing important political work.

Setbacks are treated very differently, on the other hand. The Report estimates that the COVID-19 pandemic has likely delayed global progress in addressing poverty by ‘3–10 years’. Similar brief allusions are made to the likely impacts of the Russian invasion of Ukraine. The estimate of ‘3–10 years’ (which takes inexorable progress to be the default) is based on simulations using data on school closures and comes with the significant caveat that much of the data through which the MPI is constructed was collected before the pandemic. This presentation of the impacts of the pandemic is telling — the pandemic figures here as an unforeseeable shock, disrupting otherwise ‘good work’ on alleviating an implicitly naturalized state of poverty. Yet, the estimate of ‘3–10 years’ ultimately tells us very little about the social relations through which people became impoverished in the course of the pandemic, or by the knock-on effects of war, for that matter. It also naturalizes, reifies and externalizes war and disease themselves, obscuring how global capitalism contributes both to producing these crises and structuring the distribution of the associated costs (see, for example, Brenner and Ghosh, 2022; Sparke and Williams, 2022).

Take hunger, which is particularly relevant here, as nutrition is one of the indicators of the ‘Health’ component of the MPI. There is little doubt that we are staring down a severe global food crisis, which has been prompted in part by supply chain disruptions caused by the pandemic and by Russia's war on Ukraine. Food and Agriculture Organization (FAO) estimates suggest that undernourishment has risen from 8.0 per cent in 2019 to 9.3 per cent in 2020 and 9.8 per cent in 2021, equalling roughly 150 million newly hungry people only in the period 2019–21 (FAO et al., 2022: 10). A portion of this number will unquestionably add directly to the MPI's poverty headcount as new data becomes available or will add further layers to the severity of deprivation for some of the 1.2 billion existing ‘multidimensionally poor’.9 Yet, while the crisis no doubt has been triggered by discrete events, the existence of the pandemic or the war in and of themselves does little to explain the highly variegated landscape of vulnerability to such shocks, nor does it show us how these vulnerabilities are inextricably linked to wealth and capital accumulation. Jennifer Clapp (2023), for instance, notes that one key feature linking the current global food crisis to comparable earlier episodes, notably the famines prompted by 1973/74 droughts in Sahelian Africa and the global food crisis that accompanied the global financial crisis between 2007 and 2012, is the presence of interlinked forms of concentration of control over key crops. In a similar argument, Raj Patel (2022) notes that the crisis in the global food system is deeply rooted in the control of a few key staple crops by a handful of corporations. This pattern itself, Patel suggests, is closely interlinked with colonial histories in which local food staples were replaced with a few global commodities (notably wheat and rice), organized through intercontinental supply chains. In short, spiralling prices of globally significant staple crops are very much a feature, not a bug, of our current global food regime.

The point here is that somewhat paradoxically, at the same time that the MPI brings a diverse range of deprivations like hunger directly into a poverty metric without assuming that these are mechanistically linked to income, it also militates against recognizing the wider historically embedded relations of power and accumulation within which impoverishment must be situated and understood. In this sense, it's clearly part and parcel of a global policy consensus on poverty that emerged from the 1990s onwards, as I detailed earlier. Here, Peck's (2011) more general argument that the increasingly rapid circulation of policy models and expertise effectively depoliticizes relations of poverty seems especially apposite. For Peck, such models ‘decisively pre-empt what would otherwise be variegated, locally specific debates around the causes and cures of poverty, further depoliticizing the policymaking processes through the circulation of prefabricated solutions, traveling in the disarming, apparently “neutral” and post-ideological form of evaluation technoscience and best-practice pragmatism’ (ibid.: 178). Of course, the promise of the MPI is more than one-size-fits-all — the headline claim about the value of the MPI for policy is precisely that it enables poverty responses targeted and tailored to the particular combined forms of deprivation, or ‘multidimensional deprivation’, prevalent in different parts of the world. This is a claim that can be exemplified by the focus of the most recent Report on ‘deprivation bundles’, to which I turn in the next section.

The primary focus of the MPIR 2022 is on so-called ‘deprivation bundles’, defined as ‘pairs, triplets or larger groups of interlinked deprivations that make up all or a subset of a person's deprivation profile’ (p. 6). For example, the Report points to the typical co-occurrence of four deprivation indicators of the MPI — inadequate nutrition, the use of solid cooking fuel and a lack of access to adequate housing and improved sanitation (see Table 1 above). It estimates that some 3.9 per cent of the global poor, or some 45.5 million people, the vast majority living in India, Bangladesh and Pakistan, are subjected to this particular combination of deprivations (p. 6). Another estimated 328.9 million people, mainly living in sub-Saharan Africa10 and South Asia, are subjected to all four of these in addition to others.

To a degree, this is a somewhat more formalized and intensive approach to a theme — the co-occurrence of different specific deprivations — that has been present in work on the MPI since at least 2010. For instance, in the 2010 working paper by Alkire and Santos, the index measuring various dimensions of poverty at the household level was justified precisely for its ability to identify discrete clusters of frequently co-occurring indicators. The authors of the Report note that identifying clusters in this way ‘enables us to identify different “types” of deprivations — clusters of deprivations that occur regularly in different countries or groups’ (ibid.: 8). Identifying such discrete clusters ‘can thus contribute to a better understanding of the interconnectedness among deprivations, can help identify poverty traps, and can thus strengthen the composition and sequencing of interventions required to meet the MDGs’ (ibid.).

This move towards an analysis of deprivation bundles appears to be, in part, an acknowledgement of some of the limits of the MPI as a composite index. One of the chief methodological concerns raised by critics of the MPI has long been precisely the disadvantages of compiling aggregating measures into a single index. Ravallion (2011: 237) aptly describes the dilemma using the metaphor of a car: ‘Imagine that a new car comes on the market that collapses all those dials on the dashboard into just one composite index, on which you are supposed to decide what to do (slow down or get fuel)’. The emphasis on deprivation bundles appears a kind of concession to Ravallion's argument that for the purpose of designing policies, ‘[i]t is not the aggregate index we need … but its components’ (ibid.: 240; see also Fischer, 2018: 114–17).

Although the methodological questions here are important, the political question is both more relevant to the current analysis, and perhaps ultimately more consequential. The aim of both aggregating the index and disaggregating it is to ensure ‘effective’ poverty–reduction policy making — evident even in the reference to using the MPI to ‘strengthen the composition and sequencing of interventions’ in the passage from the 2010 working paper quoted above. The MPIR 2022 makes the same stakes clear in stating that ‘looking closely at the interlinked deprivations of poor people, provides valuable insights on how to tackle multidimensional poverty … by addressing its multiple dimensions’ (p. 1). Understanding how multiple deprivations are co-located, the report states, enables the design of ‘integrated policies that can tackle multiple deprivations at once’ (ibid.). The Report is inevitably a bit vague about the more specific implications for policy, but the relevance of this ‘bundling approach’ is ultimately framed in terms that reflect and reinforce the underlying preference for targeted interventions typical of much of the global development work taking place in the last 40 years (see Mkandawire, 2005). If, as Fischer (2018: 126) noted a few years ago, ‘the MPI has been increasingly operationalised as a targeting device rather than just an evaluative tool’, the focus on deprivation bundles in the MPIR 2022 appears to fully embrace this role.

There are two ways in which the Report justifies the mapping of ‘deprivation bundles’. The first could be described as identifying those aspects of poverty of the greatest salience. To return to the example of the ‘most common deprivation bundle’ discussed above (nutrition, cooking fuel, sanitation and housing deprivations), the significance of identifying these deprivations together is that ‘[d]esigning policies that address the four deprivations in this bundle will have a high impact on poverty by bringing people experiencing this deprivation profile out of poverty and by improving the lives of millions of other poor people who experience these deprivations along with others’ (p. 6). Moreover, whereas interventions seeking to improve nutrition, sanitation and housing are often conducted separately, knowing that these deprivations often co-occur ostensibly creates an impetus for ‘multisectoral coordination’ (ibid.).

Another purpose of exploring these bundles for the MPIR 2022 authors is to ensure that interventions more effectively target different regions. The Report also considers deprivation pairs and triads — for any given deprivation score, how often it occurs in tandem with each other deprivation or pair of other deprivations. The prevalence rates of different triads of deprivation, the Report notes, vary widely between regions. Globally, about 41 per cent of poor people are deprived of access to electricity, sanitation and housing, together — roughly the same proportion deprived of adequate nutrition, housing and cooking fuel, together. However, whereas the former affects 66 per cent of the poor in sub-Saharan Africa, against 11 per cent of the poor in South Asia and 0.2 per cent in Europe and Central Asia, the latter affects 46.4 per cent of the poor in South Asia, 44.8 per cent in sub-Saharan Africa and a significant share (20 per cent) of poor people in Europe and Central Asia (p. 10). Programming thus presumably can be tailored around the specific bundles of deprivations most prevalent in particular places. Likewise, identifying deprivation bundles can help policy makers target the most severely deprived, as ‘[l]eaving no one behind means focusing on the people with the highest deprivation scores’ (p. 10). Here, the report points specifically to an estimated 4.1 million people classed as deprived on all 10 of the measures of deprivation included in the index, of whom 3.8 million live in sub-Saharan Africa. Most of the remainder, or some 210,000 ‘most-deprived’ people, hail from Sudan, which in the Report is classified as an Arab state.

Serious concerns can be raised about how effectively an index like the MPI in and of itself could or should function as a targeting tool. As a mechanism for prioritizing interventions, it is insular and recursive — by definition, it can only identify as priorities those areas of intervention that have already been selected for inclusion in the index. As a mechanism for means-testing or targeting interventions, it is similarly limited. Proxy measures of poverty generally fail to meaningfully identify people in need — Kidd et al. (2017: 1) note that a comparison to a ‘lottery’ would be a ‘reasonable assessment of [the] efficacy’ of proxy measures in targeting interventions. A focus on a handful of assets can easily result in the failure to notice changes in household incomes and livelihoods over time, or in more salient assets that are simply not measured (e.g. land and livestock). Kidd et al. (ibid.) further insist that they are in effect tools for rationing aid rather than identifying the ‘neediest’. The MPI could not possibly work as a means-testing mechanism at the individual or household level, based as it is on anonymized surveys of a sample population conducted once every few years. Instead, it promises a form of low-resolution geographic targeting of entire countries or regions based on specific ‘deprivation profiles’ or with a high concentration of overall deprivation. Even when wider questions about universalistic versus targeted approaches to poverty reduction are set aside, the limits to using something like the MPI as a targeting mechanism of any sort remain evident.

Tellingly, besides being a very high-level prioritizing and targeting exercise, the MPI can tell us little about what actual anti-poverty interventions should look like. In fact, the specific actions called for in the Report mainly relate to improving data collection so that ‘policymakers can identify emerging policy concerns, inform programme design and policy choices, forecast trends, monitor policy delivery and evaluate programme impact’. The MPIR 2022 notes that ‘[t]he irregularity of multitopic household surveys — the main tools of poverty measurement and analysis — hinders the power and potential of the global MPI’ (p. 23).

If the MPI is not in the business of circulating mutable ‘models’ of poverty reduction, its policy agnosticism nonetheless ironically creates similar political dynamics. It also, if implicitly, very much shares the emphasis that neoliberal development programmes place on targeted and projectified interventions to the exclusion of universal provision or redistributive measures. Instead of facilitating meaningful or contested discussions about the causes and consequences of poverty, the index specifies how to target palliative responses most effectively. Again, the point here is not to argue that the authors of the MPIR 2022 could or should wade into these waters directly — rather, I contend that it is politically significant that, deliberately or otherwise, the ability to do so is effectively ruled out from the start by the very nature of the exercise.

The focus of the Report on mapping ‘poverty’ in and of itself leads it to obscure or negate several important things. It contains strikingly few mentions of wealth or even inequality. Indeed, the lack of such discussions — and what is in fact discussed — is telling. It's a crude measure, but the MPIR 2022 for instance mentions ‘billionaires’ exactly once, on the first page, where the authors note the discrepancy between the limited data we have on poverty and the wealth of data on the number of billionaires in the world — ‘a jarring data inequality’ (p. 1). This observation is repeated later in the report, on page 29. The Report otherwise looks at ‘inequality’ only in the context of inequality among the poor, calculated as variance between individual deprivation scores. Of course, this is not to argue that the MPIR 2022 necessarily should focus on billionaires or the extremes of wealth — given how the index works, this would have made little sense in methodological terms. However, it is notable that such questions are effectively ruled out from the outset by the manner in which the MPI is calculated. The example of how the global food regime co-produces corporate concentrations of power and vulnerability to hunger has already been discussed above.

A relational conceptualization of poverty (see, for example, Bernstein, 1992; Selwyn, 2014) requires an attempt to map out the relations effecting wealth extraction and accumulation — in particular, in Marx's (1867/1990: 799) evocative terms, how the ‘accumulation of wealth at one pole simultaneously acts as the accumulation of misery, the torment of labour, slavery, ignorance, brutalization and moral degradation at the other side of the pole’. In this final section, I argue that in order to do this, what we need is to put forward a politics of poverty centred on mapping and confronting multiscalar relations of exploitation as a driver of poverty.

Exploitation itself is a contested term. In non-Marxist usage, the term often serves to designate coercive, deceptive and/or illegal practices, thereby placing certain practices outside the bounds of ‘normal’ capitalism. In early 2023, for instance, US President Joe Biden proclaimed on Twitter that ‘[c]apitalism without competition isn't capitalism. It's exploitation’.11 While not a rigorously formulated conceptual exploration, Biden's differentiation of ‘capitalism’ from ‘exploitation’ on the basis of the exercise of monopoly is nonetheless notable. Exploitation is often treated by others as synonymous with labour abuses and violations of basic workers’ rights. In either case, it is expressly externalized from the ‘normal’ workings of capitalism. With rare exceptions (notably Boucher, 2022), the concept is often used loosely .

The Marxist use of the term, by contrast, is useful here in that it highlights the dynamic and multiple ways in which the accumulation of capital and relations of impoverishment are co-produced. Marx (1867/1990, 1894/1991) himself admittedly uses the term ‘exploitation’ sparingly and even inconsistently across the three volumes of Capital. Subsequent Marxist scholarship sometimes follows suit, although ironically at times does exactly the opposite, defining exploitation in far too narrow and rigid terms. This is sometimes spun into a rigid reading where value can only be created through wage work in production sectors organized along industrial lines.

As Stuart Hall insists, however, it is much more useful to read Marx's emphasis on ‘free’ labour as ‘laws of tendency (and countertendency) rather than as a priori laws of necessity’ (Hall, 1980: 331). A full picture of capitalist exploitation would, at the very least, have to incorporate a diverse range of unfree modes of labour exploitation, as well as peasant and informal modes of production based on the continued nominal ownership of direct producers of the means of production. In short, there are myriad and often contradictory ways in which ‘capitalist relations of production (the accumulation and competition of capitals) can be structured in terms of the actual exploitation of labour’ (Banaji, 2011: 41; see also Hall, 1980; Mezzadri, 2021). Feminist scholars have long and rightly insisted that ostensibly productive labour depends at all points on heavily gendered, often unpaid and uncounted reproductive labour (see Mies, 2014; Rai et al., 2014). As Mezzadri (2021: 1194) argues, a recognition of the multiplicity of capitalist exploitation ‘allows us to recuperate … a broader capitalist history of the wageless across the colonial and postcolonial world, where petty commodity production, non-wage and disguised-wage labour, including forms of slave, indentured, unfree and bonded labour are the norm’.

Nor does exploitation under capitalism take place directly through labour alone. According to Marx, rents, interest payments and the like represent ‘a secondary exploitation, which proceeds alongside the original exploitation that takes place directly within the production process itself’ (Marx, 1894/1991: 745; see also Bernards and Soederberg, 2021; Harvey, 2006: 42; 2012: 29; Soederberg, 2014: 4). Extractive relations of indebtedness directly share a dynamic and recursive relationship with the exploitation of labour. ‘Usury’, Marx (ibid.) notes, ‘is a powerful lever in forming the preconditions for industrial capital’ in that it enables the formation of an ‘autonomous monetary wealth’ and, more importantly, in that it ‘appropriates the conditions of labour, by ruining the owners of the old conditions of labour’. The point is that a Marxist lens broadly applied requires investigation of the concrete historical relations through which poverty and wealth are co-created through the intersection of multiple, dynamic and overlapping modes of exploitation.

Centring the myriad relations of exploitation through which wealth and poverty are co-produced in this sense offers us a means of highlighting the links between global circuits of capital accumulation and the localized experiences of deprivation emphasized by the MPI. Ben Selwyn (2014, 2019) in particular has made similar arguments about the role of labour and global value chains in development. Against perspectives that frame employment at the fringes of global value chains as, in Jeffrey Sachs’ words, ‘the first rung on the ladder out of extreme poverty’ (Sachs, 2005: 11, cited in Selwyn, 2019: 74), Selwyn argues that poor pay and working conditions inherent in these settings in fact create new forms of poverty. For the mostly women rural–urban migrants disproportionately engaged in light manufacturing activities, increased incomes in monetary terms are accompanied by the intensification of work and the rising cost of items required for survival. For Selwyn, when we focus on the relations of exploitation accompanying the diffusion of global value chains, these types of employment start looking less like a ‘first rung on the ladder’ and more like incorporation into ‘global poverty chains’ (Selwyn, 2019: 72).

It is also worth noting that the MPI fails to directly consider work and labour questions — the type of work people do, the duration and intensity of the work, where they work, how they get there, which infrastructure and equipment they use, under which terms of employment — an omission that in fact has profound implications for many of the measures of poverty included in the MPI (e.g. health, housing access or education). For instance, Brickell et al. (2023) demonstrate how in Cambodia's garment sector, the loss of earnings due to layoffs stemming from the pandemic-era cancellation of orders from major global retailers and increasing debts, coupled with the withdrawal of state support and the offloading of responsibility for social reproductive work onto households, have culminated in a crisis of hunger. (Although, to be fair, the authors of the MPI do acknowledge the absence of labour alongside physical insecurity and health within the household as a problem and argue for the inclusion of new question modules on work — see p. 24). To some extent, measures of poverty like the MPI might well capture some elements of renewed or transformed relations of poverty insofar as they manifest as restricted access to housing, water or food. But at the same time, they cannot capture relations of ‘depletion’ (see Gunawardana, 2016; Rai et al., 2014) that often accompany the intensification of work, particularly where this occurs alongside the continuation of social reproductive labour.

In this respect, the tendency of the MPI measurements and corresponding narratives to move between the individual and household level is significant. Most of the indicators used in the MPI are household-level measures, but the poverty headcount and basic conception of poverty are very much individual-level measures. This is on one level simply a result of data availability — the underlying surveys are household surveys — but it nonetheless obscures important dynamics of exploitation, depletion and impoverishment. It is not just that access and deprivation are often very unevenly distributed within households (although this is true). More importantly, household relations structure relations of exploitation in ‘productive’ sectors. For example, Baglioni (2021), in relation to the margins of horticultural supply chains in Senegal, shows how gendered household relations can play a significant role in making women available for exploitation as cheap labour in packhouses and on farms.

Equally, paying attention to the relations of exploitation can contribute to a more substantive understanding of responses to poverty. It is perhaps one thing to know that a lack of access to electricity, sanitation and adequate housing tend to be aligned and should be the focus of interventions, but it is quite another to consider how access to these services could or should be improved. As is hopefully clear from the discussion above, the MPI is relatively agnostic about whether and how people access goods that make up their ‘standard of living’. A view of processes of secondary exploitation alongside exploitation through work is especially apposite given the particular debt-fuelled interventions that have been favoured as responses to some of the key deprivations highlighted in the MPI in recent years.12

To give one example, an increasingly common response to difficulty accessing electricity, particularly but not exclusively in rural areas, has been to promote the adoption of pay-as-you-go (PAYGO), off-grid solar photovoltaic (SPV) systems (see Quinteros and Bernards, 2023). These systems can and do directly address at least one of the deprivations (a lack of electricity) that counts towards the MPI. They might potentially also address other deprivations indirectly; for example, enabling the replacement of solid cooking fuels with an electric stove or the use of some of the consumer goods included in the measure of assets. Yet, they do so in a way that is highly contingent. Baker (2023) notes that PAYGO SPV systems have the effect of converting rural energy use into a set of financial assets ultimately grounded in new forms of consumer indebtedness. One of the results of this has been that energy is accessed only at a significant cost. Energy access through relations of indebtedness is also uneven and sporadic. Cross and Neumark (2022) show how PAYGO SPV systems in East Africa constitute an adverse ‘infrastructure of inclusion’ in which final users are governed by new circles of data, capital and debt. Thus, PAYGO business models do play a role in connecting people to electricity, but they also set the grounds for disconnecting those defaulting on agreed payments insofar as they rest on digital infrastructures that can remotely lock out or shut down systems. The possibility of remote disconnection is significant for users and businesses alike given high rates of default, on the one hand, and the notable material and social costs associated with repossessing SPV systems, on the other.

We might also note that the wider matrix of exploitations within which impoverished livelihoods are entangled often make a significant difference to the outcomes of anti-poverty interventions. For instance, improvements in access to water and sanitation that do not pay sufficient attention to localized patterns of property relations, exploitation and accumulation can have adverse effects. Likewise, slum upgrading programmes have a long-noted tendency to proceed on the (mistaken) assumption that everyone owns the home they live in (de facto if not de jure). In practice, many people in informal settlements are tenants, and upgrades to infrastructure can push poorer tenants out of their rented homes if these enable landlords to command higher rents (see Desai and Loftus, 2012).

The overarching point is that, absent a serious consideration of how to go about doing so, there is a risk that interventions targeting particular deprivations or clusters will wind up reinscribing them in a new form or indirectly intensifying pressures of indebtedness that have a bearing on other types of deprivation. Again, a relational perspective showing how processes of exploitation are connected to processes of capital accumulation is vital. Of significance here is that underlying the recourse to debt relations as mechanisms for widening access to basic services in the first place is arguably the radically uneven distribution of resources in the global political economy. The turn to various self-help or debt-based poverty alleviation interventions, and the wider ‘private turn’ in development financing in recent decades, has often expressly been justified in terms of the co-existence of restrictive fiscal constraints faced by peripheral states and deep pools of capital in global financial markets (see Bernards 2022, 2023).

Whether global poverty is rising or falling in any aggregate sense likely cannot be adequately determined and is arguably of less value than often assumed in terms of understanding how to address poverty. Aiming to understand the variegated shape that poverty assumes and the social relations that drive it, as well as the form that meaningful responses to poverty might take, could be a more helpful place to start. The MPI and other such indices no doubt are broadly useful in informing such understandings, but in their current form often serve to obscure more than they reveal.