The Politics of Knowledge and Social Cash Transfers: The Constitutive Effects of an Anti-Poverty Regime in Indonesia

Abstract

Recent decades have witnessed the globalisation of policies promoting social cash transfers as a critical instrument for poverty reduction. Among various approaches, the Conditional Cash Transfer (CCT) model promoted by the World Bank has gained discursive dominance in countries where this strategy, and its technical model for implementation, appear more attractive than competing alternatives. While research has evaluated CCT programmes and considered the politics of development that they represent in Latin America, researchers are yet to explore the constitutive effects of CCT ways of knowing and measuring poverty in the societies of rural Asia. This paper explores the consequences of CCT knowledge politics in rural Indonesia. It argues that CCT practices of knowing and measuring have paradoxical effects. The programme makes direct payments to millions of impoverished households, producing well-documented patterns of inclusion and advancement. Yet, CCT knowledge practices involve simplifications and generate significant mis-targeting, eliciting a never-ending repair process among state actors, local leaders, and communities. This metricised knowledge system depoliticises political questions of distribution. It conceals alternative ways of knowing and addressing poverty, producing an order of entitlements somewhat at odds with established community logics of inclusion, while provoking a local politics of distribution.

Key Words:

• conditional cash transfers
• food security
• Indonesia
• poverty
• rural development
• social assistance

Since Indonesia began its social cash transfer programme in 2005, it has expanded to become “the second largest program” of this kind in the world (Tempo, March 26, 2017). The “Family Hope Programme” (Program Keluarga Harapan or PKH) represents Indonesia’s first comprehensive attempt to cover poverty on an individual basis, by targeting the poorest 10% of society. In 2016, 6 million households received PKH via a series of initially irregular payments. In 2018, the programme was improved and extended to 12 million families (an estimated 40 million people) who received 1.8 million Rupiah (US$133 at 2018 prices) per year. In 2019, yet another massive increase in payments was initiated to provide regular cash transfers to 15 million households. During the COVID-19 pandemic, the PKH programme was one of the backbones of the government’s mitigation and support programmes (World Bank 2021). It represents Indonesia’s critical policy response to poverty and inequality (Suryahadi and Al Izzati 2018, 201–202). The national press and various evaluation reports praised it and supported its expansion. Research has found conditional cash transfers positively impact school enrolment, access to health care, maternal mortality, and child nutrition (see, for example, Cahyadi et al. 2018; Satriawan 2016). However, regional newspapers have discussed a range of other issues. Headlines include: “PKH assistance considered to be mis-targeted,” (Serambi, August 14, 2018) and “Demonstration by Housewives and Elderly Criticise Mis-targeted PKH Assistance” (Pikiran Rakyat, March 14, 2019). Articles reported the use of out-of-date data, unmerited choice of beneficiaries, the inclusion of the non-poor and exclusion of the poor, widespread dissatisfaction, jealousy, and contention within communities, stigmatisation of the poor, and demonstrations by villagers and even village heads (McWilliam et al. 2023, 73–75). Official reports also note a high exclusion rate (TNP2K 2018, 2).

In parallel with Indonesia, recent decades have witnessed the globalisation of policy thinking regarding the centrality of social cash transfers in the management of poverty. While an extensive range of policy approaches have emerged, the Conditional Cash Transfer (CCT) model promoted by the World Bank became an ascendant instrument in many countries (Leisering 2019, 3; Peck and Theodore 2015). Over 130 countries have implemented CCTs including many in Latin America, Indonesia, and the Philippines (Olivier de Sardan and Piccoli 2018a, 35). Project evaluations and the economic literature have praised the model’s contribution to increasing household consumption, human capital development, and improving nutrition. Research has also pointed to persisting problems of mis-targeting, unfair allocation of benefits, an inability to overcome social exclusion, and a general failure to address the structural causes of poverty (see Ladhani and Sitter 2020). Commentators have noted that the model coincides with a global economic context dominated by market-oriented policies, the financial development for the poor, and (before COVID-19) a quest for fiscal efficiency (Ferguson 2015). Yet inequalities have grown, and governments have yet to provide a structural solution for the intertwined issues of poverty, marginalisation, and exclusion (Olivier de Sardan and Piccoli 2018a). Despite such criticisms and a recent move away from CCTs in Mexico, they have remained popular and influential (Beck, Regidor, and Iber 2020). During the COVID-19 crisis, governments used CCTs to curb the harmful effects of unemployment, quarantine measures, and rising poverty levels (Busso et al. 2021).

Identified by Leisering (2009) as an “entitlements revolution,” these cash transfer programmes produce a new politics of distribution (Ferguson 2015). As the Indonesian case shows, social assistance programmes create new modalities of distribution, presenting new opportunities for making claims on the state, while opening up new sites of contention and possibilities for political mobilisation over “who gets what, when and how” (Lasswell 1936). Field research has yet to consider fully the politics of distribution provoked by CCT programmes as they move into Asia.

At the same time, flourishing policy models depend upon a politics of knowledge, that will ensure that, amongst competing knowledge claims, a particular approach and its technical model for implementation appears more reasonable than others (Peck and Theodore 2015). This knowledge politics includes repair processes involving translation, revising, or reworking to facilitate the acceptance and application of new policy models (Hasan et al. 2019). Especially in poverty eradication and social welfare, the ascendancy and hegemony of a model denotes a significant ideational shift (see Cook and Kabeer 2016; Somers and Block 2005). This pre-eminence occurs through the application of travelling technologies – specialised knowledges, models, and techniques that are transferred and mobilised across institutional environments – whose effects may be less visible, overlooked, or unchallenged (Li 2007; Nooteboom and Rutten 2012). Policy discussions that depend upon econometric logics seldom consider the links between ways of knowing, justifying, and implementing particular policy designs and the processes of contestation, adaptation, and transformation that they provoke (Hasan et al. 2019; Peck and Theodore 2015). Yet, they deserve critical scrutiny.

This article borrows the notions of translation and repair from Science and Technology Studies (STS). STS scholars suggest that technological artefacts (such as CCT knowledges and practices) work as “texts” or “scripts.” They rest on quantitative measures and technical norms, inscribe visions, scripts, or scenarios, and give rise to specific organisational forms. Yet, such technologies require translation into a context. This involves using processes where users accept, negotiate, or reject the roles and prohibitions set out by programme designers. In other words, programme logics with embedded ways of knowing, measuring, and allocating resources are not “docile objects with fixed attributes,” but rather they have considerable “interpretive flexibility” (Lu and Qi 2022, 8). As a result, people, power, technology, and resources are connected and mutually influence each other to exert control. They appear as assemblages, “constellation[s] of heterogeneous elements and forces that in coming together are consequential” (Guthman 2019, 17). In STS, often, political economy approaches and issues of inequality are poorly addressed, and people’s agency may be underestimated (see Guthman 2019). In this article, interests, power differences, and agency are central, and the research draws on environmental and rural development literature, marrying STS insights with a political economy approach.

Intermediatory actors, such as the World Bank and social welfare field officials, translate technologies, such as CCT designs, implementation models, and related technical information, during processes that are socially shaped and affected by human agency. Just as the translators of a text anticipate substantial reinterpretations of the source text in constructing a version in another language, translation of travelling technologies can be unpredictable, non-linear, and involve slippage. Actors aim to rework dysfunctional processes or failures. This “repair” involves learning, innovation, rethinking, and even repurposing programme software in a creative process of revising or working around dysfunctional knowledge and methods (Lu and Qi 2022). Such translation and repair processes significantly shape how CCT knowledge affects social entitlements and mutates during application. Moreover, as discussed below, these politics of knowledge serve the ideological and material interests of specific actors and their projects and provoke responses. Crucial to the analysis is an understanding of the implicit ideologies of CCTs and accompanying technologies, the logic of local social relations, and the reactions of local actors (recipients, local government actors, and religious/village elites) to the deployment of these knowledge practices.

Alongside the positive evaluation literature on CCTs, researchers have analysed the rationale, politics, and everyday operational problems presented by social cash transfers in Latin America (Ladhani and Sitter 2020). A recent edited volume discussed how target populations perceive CCTs in Latin America and Africa (Olivier de Sardan and Piccoli 2018b). Researchers have unpacked the politics of knowledge that underpin CCTs, by focusing on the national-international scale at policy and conceptual levels (Ferguson 2015; Fischer 2018; Peck and Theodore 2015). Building on this work, there is a need to explore more fully how CCTs mutate in diverse local contexts particularly as they move into Asia. How does this knowledge reshape the policy field, generate socio-bureaucratic adjustments, and provoke a politics of distribution that lead to processes of contestation and reconstitution? How do these knowledge politics cause constitutive effects (across scale)? The term “constitutive effects” seeks to capture the way knowledge modalities associated with the governance of poverty (such as the methods of defining and measuring poverty) help determine social practices and experiences (Dahler-Larsen 2012, 173).

This article contributes to the broader literature on social assistance by showing how this politics of knowledge produces its constitutive effects. Following others, this article analyses how knowledge politics underpins a prevailing view of poverty, setting out how it should be measured and addressed and how social assistance should operate. As particular forms of knowledge production gain ascendancy, policy specialists with the required analytical (econometric) training gain authority and power within this particular policy field. Taking this further, the links between the knowledge regimes that govern CCT modalities and the processes of enculturation of CCT knowledge and practices are examined, as well as the interactions and mediation processes at lower government and community levels. Also analysed are the practices of translation, repair, mutation, and everyday politics that emerge as the CCT model is applied in rural Asia. From this perspective, the well-documented problems associated with CCTs are shown to be more than “implementation problems.” The article suggests that they are co-produced as the knowledge regimes meet local contexts. This article advances an understanding of the effects of CCTs that indicates that, instead of merely adjusting procedures or design features of the CCT package to “fix” the programme, advancement to a more equitable, accountable, and inclusive social assistance system will require a transformation in the governing knowledge politics.

Earlier research into knowledge politics developed several insights. First, “the ways in which we think about and represent reality” are intimately linked to “the ways in which it is acted upon and governed” (Lövbrand 2011, 227). Hence, knowledge and political orders co-produce one another (see Allan 2017). Global policy models emerge from epistemologies and institutions that privilege specific ways of knowing and acting over alternatives. The forms of knowledge that underpin techniques for addressing poverty alleviation provide the overall logic for governing poverty (Methmann 2011). Second, while blueprint models have a particular attractiveness, with the promise of positive outcomes, best practice learning, and quick implementation, they involve hidden biases and rest on the normative assumptions of specialist knowledge. Third, successful policy technologies can marginalise alternative perspectives, including those required for effective interventions and negotiating pathways to better outcomes (see Scoones 2016). Hasan and colleagues (2019) and Peck and Theodore (2015) have shown that travelling policy models can work to support policy coalitions, but their knowledge regimes can simplify complexities and hence struggle in the face of the tangle of political, social, historical, and cultural factors that shape outcomes as discussed by Scott (1998) and Vetterlein (2012). This is because, according to Hickey (2012), they can overlook the dynamics of the local political economy and the logic of locally embedded institutions, including cultural understandings of particular circumstances, social expectations, and notions of reciprocity (see chapters in von Benda-Beckman, von Benda Beckman, and Marks 2000). Finally, as travelling policy models confront bureaucratic and local political obstacles, they mutate in surprising ways, provoking a new politics of distribution around how to distribute resources, who is entitled to receive them, and why (Ferguson 2015, 1–35). This process generates local practices of repair. Simultaneously, by transforming political questions of distribution (who gets what, when and how) into technical problems of programme implementation, these policy technologies can mystify and obscure the underlying politics of knowledge and the major ideational shifts taking place in the field of anti-poverty policy and welfare.

This article draws on such insights, focusing on the following questions: how do the predominant politics of knowledge of CCTs work and what are their constitutive effects? In answering these questions, we consider how these policies generate specific “implementation problems,” provoking a local politics of distribution and repair, as actors translate, adapt, and change programme logics to avoid adverse effects or to pursue their own interests. This article argues that ways of knowing, measuring, and implementation found in the CCT model have several critical constitutive effects. First, the CCT design has introduced new metricised ways of thinking about poverty. These knowledge systems seek to render poverty governable, standardising the social, turning it into legible and administratively implementable formats. As this knowledge system combines simplification with over-complex, centralised delivery mechanisms, it generates unfair distribution and mis-targeting problems. Second, CCT policy implementation practices elicit a never-ending repair process as state actors attempt to resolve the issues of the mis-targeting and misattribution of funds, for instance, by bringing in technologies and processes from elsewhere in an endless process of recontextualisation. Third, CCT outcomes produce new patterns of inclusion/exclusion, that create orders of entitlement at odds with local distributional logics. Together, these problems create a locally contentious politics of distribution. As poverty status becomes a resource that benefits those so classified, CCT practices provoke strategic behaviour among local actors. Local actors seek to ameliorate adverse local effects, which generate an informal (and largely undocumented) reworking of CCT processes. Hence, while this politics of knowledge makes redistribution, from an implementation perspective, highly technical, it turns political questions into specialised implementation problems, depoliticises them, and displaces other conceptualisations of how to address poverty. Yet, somewhat paradoxically, as the same politics of knowledge creates new exclusions, it provokes a contentious local politics of distribution, causing social unrest, and an erosion of mutual trust and community cohesion.

This study is derived from findings collected during intermittent periods of fieldwork conducted in Indonesia during 2016–2019. The researchers are independent and had no affiliation with the programmes involved in this research. The research followed the construction of Indonesia’s social registry (the Integrated Database Update, PBDT) and continued over the period when district governments became responsible for data collection. The researchers gained acceptance from communities and key informants through a long period of engagement. The names of villages are not provided for ethical reasons.

This research applied mixed methods. It traced the role of policy knowledge across sites or scales, using open-ended interviews with officials at the national and district levels, studying policy documents and CCT implementation processes, following up with in-depth inquiries in selected fieldwork locations. The study used a comparative case study approach, selecting cases to identify contrasts and convergences across different contexts, thereby enhancing the analytical generalisability of our results. Fieldwork focused on villages in four sub-districts: two villages in Java in Bantul and two in Aceh in Aceh Besar and Aceh Utara. These areas were selected because, although they have similar rice paddy production systems, they present contrasting institutional and social contexts in the outer and inner islands of Indonesia.

The research confirmed its findings by repeated observations, triangulation, saturation, management of bias, and a search for counter examples. Livelihood dynamics were assessed using an ethnographic approach in the two study provinces. This assessment included a wealth ranking exercise using a simple participatory tool that engages key village actors in exploring poverty dynamics and articulating local concepts of poverty (Krishna 2006). Village focus groups ranked households against this emic concept of poverty and charted their movement vis-à-vis this definition. Researchers conducted a food security analysis (using the FANTA Household Food Insecurity Access Scale) to understand experiential aspects of food poverty (Coates, Swindale, and Bilinsky 2007). Systematic livelihood surveys were used to provide descriptive statistics, comparing these with official state statistics (for example, regional stunting poverty and the National Socio-Economic Household Survey [SUSENAS] data). The study compared state-targeted benefit allocation with community poverty rankings to assess how effectively state-based and informal mechanisms address vulnerability. Through triangulation between CCTs’ beneficiary lists, community-based poverty ranking exercises, zakat alms distributions – an obligatory religious responsibility of Muslims to give alms to the poor – in Aceh, and Javanese mutual help systems, this study sought to understand patterns of inclusion/exclusion that emerged from the CCT programme. The research pursued a comparative approach to understand both unique and common causal processes shaping the experience of cash transfer programmes, applying the same methodology in each case. This longer-term comparative approach enabled the study of the arrival of CCTs, early adoption, and evolving local perceptions of unfair beneficiary selection and mis-targeted distribution in the context of an emic understanding of the agrarian basis of poverty, and cultural and religious perceptions of deprivation and entitlement (Nooteboom 2019; McCarthy 2020).

The following section discusses the connections between the politics of knowledge that shape CCTs and the broader politics of development in which they are embedded. Then, drawing on these insights, we briefly consider how the CCT knowledge regimes became ascendant in Indonesia, expanding the scope of administrative power and establishing a depoliticised matrix of expert knowledge. The article then considers the view from below, by considering how the CCT model works, and analysing the processes of enculturation and contestation provoked by the CCT regime delivery in the case study villages. The study explores how the CCT politics that emerged locally as CCT knowledge and practice have generated distributional effects and confronted local socially embedded notions of entitlement and poverty. Finally, the article discusses the constitutive effects of cash transfers and the complex politics of translation.

The Politics of Knowledge and CCTs

As social assistance has gone global over recent decades, a consensus has emerged regarding the virtues of social cash transfers. Various approaches have emerged (Davis et al. 2016). However, the “Washington Consensus” agencies have focused on one model: CCTs. Researchers have described this as a liberal-residual concept of social welfare where modest means-tested benefits are confined to the very poor as a last resort, emphasising targeting, poverty reduction, human welfare investments, and conditionalities (Leisering 2019, 3–8).

The development politics behind CCTs became apparent in Mexico’s paradigmatic Oportunidades-Prospera model. This programme replaced the notion of an unconditional entitlement based on need and introduced CCTs for people experiencing poverty, then gradually substituted food subsidies and help-in-kind with food (e-voucher) transfers for the bottom two deciles using the same targeting methodology. Planners aimed to cement a social consensus around structural changes, rolling out a compensatory CCT programme alongside unpopular macro-economic changes, thereby seeking to demonstrate that disliked structural reforms would not increase poverty. Beneficiaries received CCTs provided they met certain requirements including child attendance at school and involvement in health programmes, thus linking social cash transfers to the poor with human capital development (Peck and Theodore 2015, 85–129).

CCTs have the characteristics of “silver bullet fixes” – they appear to provide a solution to wicked social problems. CCT framing uses arguments that appeal to policymakers. CCTs mobilise the idea that focusing on children, nutrition, and education is the most effective way to break the cycle of inter-generational poverty. CCTs directly offer income support to low-income families, ostensibly bypassing the bureaucracy (Lomelí 2009). The concept of conditionality fits with liberal ideas of responsibility and mutuality. The focus of CCTs on women, children, and pregnant mothers sits well with the widely accepted assumption that mothers spend money more appropriately than men. CCTs are seen as non-paternalistic because recipients can spend the assistance according to their preferences (see Leisering 2009). As Lomelí (2009) points out, many of these propositions lack evidence or are contested, yet these programme attributes have helped CCTs become a popular anti-poverty instrument. These ideas constitute a specific politics of knowledge around the construction of poverty.

The “scientifically calibrated model” of CCTs establishes another crucial element that has shaped the global success of CCTs, providing the case for reform by specifying “the prototypical machinery of policy” articulated in persuasive technical concepts (Peck and Theodore 2010, 206). The model relied on econometric methods that entailed quantification in all steps, from programme design to application. This technical approach appeared politically neutral and “objective,” with the impression of scientifically proven know-how underpinning its legitimacy. It fitted with a shift to “post-clientelist approaches” where programme implementation worked according to a system designed to avoid elite capture and clientelism by applying methodologies beyond the influence of corruption, prejudice, and the arbitrary power of elites (Maiorano 2014, 96). This involved using systematic knowledge systems that aimed to allocate benefits according to laws, rules, and policy criteria established along Weberian lines that work in an impersonal manner using a disciplined and impartial bureaucratic apparatus guided by top-down control (see Manor 2016).

Under its technical guise, this knowledge system constitutes a form of politics because it sustains a set of interpretations that shape a field of inquiry and establish a set of development practices. The CCT regime specifies how the poor are to be governed through demarcating, measuring, and categorising poverty. This entails the framing and construction of poverty as an object of governance by rendering poverty governable through targeting, steering, and controlling the behaviour of actors and institutions in an effort to realise specific policy goals (Dooley and Gupta 2017).

The literature suggests that for objects that are contested and difficult to measure (such as poverty), how those objects are quantified shapes the realities on which they focus (Merry 2011). The uses of indicators and measurement produce “a more or less orderly view of an otherwise disorderly reality” (Dahler-Larsen 2014, 976). By defining the social realities of which they are a part, measurement also guides practical action. As decisions are made based on statistical information, understandings and issues less amenable to quantification tend to be left aside, leading to simplifications. Moreover, biases based on normative assumptions embedded in specialist decisions are concealed.

The social constructivist perspective applied here suggests that ascendant knowledge systems and ways of measuring have constitutive effects on whole policy fields (see, for example, Andersen 2020; Dahler-Larsen 2012, 2014). They underpin the dominant or paradigmatic understandings of policymakers, by supporting particular policy and programme logics and thereby shaping implementation, including the relations and interactions of those responsible for carrying them out. However, as those in the policy field retain agency, and programmes encounter messy social relations as they travel, they also provoke responses, including the processes of translation, repair, and the everyday village politics of social assistance discussed below.

Yet, practical knowledge of political economy drivers, incentives, and interests that shape how governance regimes function is necessary for effective interventions and for negotiating pathways to better results. For, while blueprint designs may overlook bureaucratic motivations and local political cultures, as we will discuss below, they may ultimately shape how programmes work on the ground.

The Politics of the Ascendance of CCTS in Indonesia

Somewhat in parallel with Latin America, Indonesia’s CCT approach emerged after an economic crisis and structural adjustment programme under the tutelage of the World Bank (Kwon and Kim 2015; Sumarto 2017). As in Latin America, the rolling out of CCTs accompanied macro-economic changes as planners worked to cement a social consensus around structural changes by allowing unpopular macro-economic reforms to be paired with policies to assist the very poor (Peck and Theodore 2015).

Yet, in Indonesia, domestic structural and institutional factors, working with domestic interests and experiences, shaped the ascendancy of CCT knowledge. The CCT approach fits nicely with the political needs and interests of a range of actors. Presidential candidates (President Yudhoyono then Joko Widowo) seeking election could claim credit for expanding a programme that offered money for the poor, while, at the same time, using the programme to reduce poverty statistics and attract poor voters, as well as forging political ties with poor households and accumulating political capital (Sumarto 2017). Furthermore, broader structures of power and interest shaped the emergence of these specific welfare settings. While the union movement had advocated for an expansion of social assistance policies, in the absence of large-scale social mobilisation from below, policymakers could be cautious. Avoiding a broader, more universal approach, policymakers can favour restrictive policies that provide small benefits to a limited pool under a top-down state mechanism (Garay 2016). The coalition of actors that dominates Indonesia’s political settlement could resist wide-ranging distributional reforms while supporting residual approaches that focused on the very poor, human resource development, political stability, and economic growth (Rosser and van Diermen 2016). In addition, advocates of the CCT model mounted a persuasive technical argument (see Kwon and Kim 2015; McCarthy and Sumarto 2018). This “best practice” model promised a relatively cheap and efficient means of lowering poverty statistics and addressing difficulties in Indonesia’s social programmes (Hastuti et al. 2008, 32). The CCT model targets the “deserving” through a highly technical process that moves the allocation of social benefits beyond elite capture and clientelism.

Indonesia’s CCT programme Program Keluarga Harapan (PKH) or “Hope for Families” constituted a paradigm shift in social welfare thinking. There was a change from state-led development focusing on production and project-orientated assistance to affordable social investments that support growth and stability (Rosser and van Diermen 2016, 346). This ideational shift is a shallow ideology of inter-generational social investment that supports simple cash transfers to the very poor. While many CCT programmes in other places are not sustainable financially, in Indonesia, due to their benefits in elections and the fit with a wider politics of development, CCTs are likely to endure for some time.

Processes of Enculturation and Contestation

New Metrics: The Roll-out of the Knowledge System

Implementing the CCT model required institutionalising a standardised model using centralised criteria and indicators. Although the programme was rolled out slowly, with policy corrections over time and many policy adjustments taking several years to be fully implemented (see Table 1), there were three critical devices at the centre of the CCT knowledge system. First, following the Latin American example, the system uses a proxy means test (PMT) or econometric targeting to determine which households are the most deserving of assistance (Brown, Ravallion, and van de Walle 2016). Second, the programme developed a social registry (Basis Data Terpadu or BDT) based on village surveys that focused on the poorest 40% of the population. The surveys were designed to ascertain the status of households by gathering accurate statistics regarding demographics, human capital, type of housing, durable goods, and productive assets. Third, the econometric targeting applied an algorithm derived from statistical models using proxies to give each household a score, calculate a household’s level of welfare relative to others, rank them from poorest to most prosperous, and generate a list of beneficiaries (Kidd Gelders, and Bailey-Athias 2019, 2).

Table 1. Evolution of social assistance knowledge management

2005	Central Statistics Agency (BPS) began developing an integrated database for targeting social protection/poverty alleviation for the Direct Cash Assistance (BLT) and the PKH; PPH piloted in seven provinces.
2011	BPS oversaw data collection for 40% of households. National Team for the Acceleration of Poverty Reduction (TNP2K) began to rank the data using the PMT to create a Unified Database.
2015	BPS updated the data under the Integrated Database Update (PBDT) activity using Public Consultation Forum (FKP)
2016	The Ministry of Social Affairs (MenSos) began managing the data with Regional Governments now to be given responsibility for updating it.
2017	MenSos developed the Next Generation Social Welfare Information System Application (SIKS-NG), a data management system. Determination of integrated data now to take place twice a year.
2019	Mensos changes all the nomenclature; the system now to be called integrated social welfare data (DTKS).
2020	DTKS data is to be determined 4 times a year. Regional governments to use the SIKS-NG Application.

Source: Kemensos (n.d.).

PKH targets impoverished families in the bottom 10% who meet five criteria: pregnant women, children under 5 years of age, households which have school-going children (up to senior high school), elderly people (now over 60), and disabled people. As in Latin America, transfers are conditional: beneficiaries must send their children to school, appear at regular health checks, take food supplements, get vaccinated, and engage in family development skills workshops. By 2018, PKH had targeted recipient households with an annual income of 2 million rupiah (about US$148), covering 15 million families (Kemensos n.d.). In 2022, CCTs were seen as pivotal in fighting the social consequences of the COVID-19 crisis.

While poverty data often take on an “objective” abstract or disembodied statistical character, field research revealed the messy, relational processes through which statistical data emerges. Interviews with officials during the data collection process – the PBDT – revealed the local dynamics that shape how surveys occur. In Aceh Besar, enumerators undertaking surveys were required to revisit remote hamlets several times if people were away. Enumerators were paid a flat fee for each village and transport costs, and they had limited time. When enumerators took shortcuts, this led to errors, including the exclusion of poor households. During surveys for updating the data, enumerators also faced criticism from those excluded from the earlier beneficiary list. The task became too onerous at a certain point, and surveyors stopped going door-to-door. The sub-district Social Welfare Officer (TKSK) recalled seeing the enumerator sitting with the village head filling out survey forms. In Aceh Utara the village secretary provided information, and the enumerator completed a large set of forms. In Central Java, enumerators worked assiduously, but due to the high numbers of potential informants, they could not survey all houses. In Aceh some informants were critical, suggesting that surveyors lacked the competence, authority, or discernment to do the survey well, even if recruited from the village. In the Aceh Besar village, the village head selected a young unmarried woman to carry out the survey. A villager complained that she was “too close” to the village head: “She accepts it when he is proposing several people” (Interview, Aceh Besar, February 2016). The role of key villagers in guiding the pre-list and shaping how forms are completed led others to suspect that this increased the likelihood of individuals being included or excluded. One villager questioned whether she was confident enough to press older, recalcitrant male interviewees to answer questions truthfully or otherwise to induce them to provide accurate information. In Java, villagers were not aware of being surveyed at all. In the study village lists had been compiled by the hamlet head. Later, villagers did not comprehend what had happened, suspecting that enumerators had just filled in lists based on walkthroughs and available village statistics that are often considered unreliable.

Researchers have suggested that measurement processes lead those governed by them to change their behaviour to enhance their scores (Merry 2011, 88). Respondents wishing to maximise the chance of getting listed as poor and receive programme funds could hide assets. One beneficiary in Java even owned a car, which he did not reveal to the surveyor, according to his neighbours (Interview Imogiri, Bantul, April 2019). Surveyors found it challenging to extract accurate information about the status of land and other vital assets. While more savvy villagers could field questions well, the poor might inadvertently imply they owned assets they rented or managed.

There is a wide acknowledgement of simplification issues. Given the problems of measuring household income and consumption, the programme selected indicators for surveying based on the ease with which enumerators could observe them. Local officials recognised the challenges of differentiating the poor through a survey model that focused on using assets as a proxy for welfare. According to one village official, “People may have a poor house but extensive gardens or rice paddy; alternatively, a house may be tiled and well-roofed, but the family may have a low income” (Interview, Aceh Utara, May 2017). Moreover, some indicators are misleading. For instance, ownership of a motorbike was a proxy for household wealth, and the survey asked whether a household owned a motorbike. However, families held motorbikes on credit in many cases, and old, second-hand motorbikes are essential for the livelihoods of wage labourers and commuters. District informants argued that as the survey provided a superficial understanding of household welfare, the process required a more complex analysis which would be difficult for a simple survey to capture.

The literature has long questioned the accuracy of this survey approach. Households accumulate assets from past work, and assets are “often a poor indicator of present income” (Kidd, Gelders, and Bailey-Athias 2017, 12). As livelihoods are highly dynamic, the accuracy of PMT assessments degrades rapidly. Kidd, Gelders, and Bailey-Athias (2017, 18) concluded that “the majority of households identified by the mechanism as being in the poorest 20% of the population are, in fact, in the poorest 40%.” These authors note that “when a programme is targeted at the poorest 10% [such as in Indonesia], the in-built exclusion error tends to be around 60%. In other words, the scheme excludes around 60% of the target group” (Kidd, Gelders, and Bailey-Athias 2017, ix).

The research conducted a wealth ranking exercise using community-based poverty criteria, and Krishna’s (2006) stages of progress methodology. This involved asking key village informants to define what it means to be a poor household in local terms and identifying a socially constructed poverty cut-off – the point at which the community no longer considers a household poor. A group of local people then assessed households’ relative assets and capacities. According to this village-based wealth criteria, key informants ranked households into poor/non-poor categories. They then labelled their status positions over time against this village-identified poverty cut-off. While it is challenging to do wealth-ranking exercises well, this system enabled the development of a proportional randomly stratified sample, and the selection of a 40-household sample in each village followed by a household-based livelihood survey to confirm this ranking. The study compared the list of villagers considered poor by village standards with the list of PKH recipients.

In Imogiri district, Central Java, the studied village with about 10,000 inhabitants, 23% of all households received PKH in 2018. The wealth-ranking exercise showed that the programme aided only 44% of the potential PKH recipients – those who were both poor and who meet the programme prerequisites. The programme allocated 23% of benefits to non-poor villagers, mostly households just above the poverty cut-off point. While the PKH programme does not explicitly aim to assist the nutritionally insecure, food security surveys revealed that approximately 60% of the PKH recipients were food secure and not extremely poor. However, the programme reached only half of the severely food insecure households. This suggests that the PKH distribution missed many severely food insecure households – the poorest of the poor – while reaching only a third of the moderately food insecure. The proxy means testing thus overlooked many of the neediest.

In Aceh, the comparison of the community poverty-ranking exercise with the PKH recipient list showed that 23% of those classified as poor and meeting PKH criteria received CCTs. However, 44% of those receiving benefits were considered non-poor by the village wealth-ranking exercise. A comparison was also made with zakat distributions. Zakat relies on religious values and binds those involved in distribution through moral accountability (Pardiansyah, Abduh, and Fakhrudin 2021; Yaumidin 2011). Zakat is a non-state process that provides only limited social assistance to low-income households, usually rice rather than money. The distribution of collected funds or rice packages in rural Aceh occurs at harvest time and the end of Ramadan. Analysts often compare zakat alms practices with state social assistance programmes (such as PKH), and zakat has also inspired social assistance reforms (see Halimatusa'diyah 2015). In allocating zakat, a committee organised under the local mosque selects poor recipients using similar emic concepts of poverty to those applied in this research, differentiating poor (miskin) and indigent (fakir) recipients. Although zakat and CCTs differ, they both distribute resources to impoverished villagers. Nonetheless, it is possible to compare how the zakat process uses locally embedded ideas of poverty and vulnerability to choose those deemed worthy of assistance with the outcomes of PKH’s techno-managerial process of measuring poverty and allocating benefits.

Comparing the result of the SOP with those receiving alms under Islamic practices, the study found that the zakat provided benefits to 37% of households, comprising 70% of those identified as poor by the village wealth-ranking exercise. When the study compared the PKH list recipients with the food security survey results, it found that 17% of food-insecure households in the sample obtained CCT benefits. In comparison, 56% of food-insecure families received zakat. This represents a severe mismatch between econometric targeting and local conceptualisations of poverty.

While these descriptive statistics form a small sample and cannot be generalised to the whole population, they point to enduring patterns of exclusion and inclusion confirmed by other research and the media (McWilliam et al. 2023). Analysts have concluded that econometric targeting tends to overestimate living standards for the poorest and underestimate them for the richest (Brown, Ravallion, and van de Walle 2016). One study concluded that most households “identified by the mechanism as being in the most deficient 20% of the population are, in fact, in the poorest 40%” (Kidd, Gelders, and Bailey-Athias 2017, 18).

Translation and Repair: Local Workarounds

The state and its donor partners have continued to engage in a series of repairs to sustain the CCT model (see Table 1). The programme imported “best practice” technologies from elsewhere, and constantly revised the programme. For instance, World Bank research found that involving the community in targeting improves community satisfaction (Alatas et al. 2013). In 2015 the programme borrowed a hybrid approach from Mexico, which combined elements of community-based targeting with econometric PMT targeting. Village leaders would participate in a process known as updating the single registry for social aid recipients (the PBDT), which involved community-level deliberations (the FKP) verifying recipient pre-lists before another round of surveying and econometric targeting.

As the problem of what villagers considered unfair beneficiary selection and mis-targeted distribution remained unresolved, the programme underwent another set of changes. As policymakers continued to repair the mechanisms for collecting and verifying data, after 2017 they introduced a Verification-Validation Mechanism for Social Welfare Information System (SIKS-NG), now under the Ministry of Social Affairs (see Table 1). Once again, implementing a more technically advanced system for re-surveying recipient populations promised to solve these problems. Nevertheless, rolling out into remote districts took several years.

Just as the state programme repeatedly attempts to address these “implementation problems,” local officials and village leaders embark on a parallel process of translation and repair as they take initiatives to adjust to or work around issues posed by programme mis-targeting.

In all studied villages, unfair beneficiary selection and mis-targeted distribution corroded community trust towards the village government. In the eyes of excluded villagers, the village head and village officials were to blame. Villagers who deemed themselves deserving, but who had been left off the list, complained about the process. Jealousy abounded. During the early phases of the programme when BPS collected the data, district statistics’ officials reported villagers protesting outside their offices and coming to their houses to complain. In Aceh, officials told stories of excluded people physically threatening village heads and making allegations of corruption to higher authorities.

From 2015, village consultation meetings and district government officials had a more significant role in overseeing the process. Discussions of social assistance commonly blame local clientelism for challenges faced by the targeting system. Yet to what extent were village leaders to blame for mistargeting? During interviews in Aceh, village heads said they enlisted people they held as “deserving” of assistance after considering poverty levels, assets, land, housing, and the number of children. In the case study areas in Java, village officials expressed that they sincerely tried to include the poorest of the poor although, in all villages, preferential treatment of neighbours and relatives was also reported. In Aceh officials also discounted clientelism. Field-level social affairs’ officials, overseeing several social programmes across dozens of sub-district villages in Aceh, noted that only a small percentage of villages experienced a chronic nepotism problem where village leaders included relatives and others close to them. Village heads pointed to ways in which village accountability mechanisms shaped their behaviour. One village head admitted that “family factors” did play a role but were scarcely significant. He argued that “five from a hundred choices would have a family nuance” (Interview, Aceh Besar, March 2019). While it is undoubtedly possible for village heads to enter some family members onto a list, such recipients would need to be from modest households to avoid drawing too much criticism. A village head who flagrantly abused his position would lose respect and authority.

During fieldwork, the outlines of various village workarounds to reduce pressure on village heads gradually became apparent. First, blame-shifting proliferated. When disgruntled villages pressured village heads and visited sub-district and district offices to lobby for their inclusion on the beneficiary list, officials and facilitators explained that the data came from a central agency (TNP2K and later MenSos). They explained that nothing could be done until the list was reviewed and renewed. Although removing names from the list was technically more straightforward, officials feared a backlash from the excluded, and generally, they did not dare remove names. Most of the work went into resolving the issue in the village, with endless deliberations and explanations to village members of their inability to change name lists: “The list has been decided upon in Jakarta, we cannot change it” (Interview, Aceh Utara, March 2019).

While the programme provided village-level deliberative meetings – the FKP – to fix beneficiary lists, the convoluted and centralised social welfare system proved to be overly complex, error-prone, and subject to data management problems (see Katiman 2023). Despite villagers repeatedly proposing changes to the list of beneficiaries, village administrators saw the same unchanged list of recipients returning from the Social Affairs Ministry. They became disillusioned with trying to change the list. In the Aceh, village heads deemed the deliberative meeting unnecessary, and sidestepped it. Arguing that they knew who was poor, they avoided social tensions and maintained their reputations by shunning complaints from those excluded by avoiding meetings to discuss the list and sending the same data back to the central government. They sought to evade discontent by allowing the government programme to continue as usual. In Central Java, deviating from government instructions to organise the deliberative meetings was unthinkable. According to the instructions, village administrators held meetings where hamlet level elites discussed the programmes and their problems, but these meetings were pro-forma meetings, and nothing changed. The leaders re-submitted the list of beneficiaries without changing it.

Village administrators sought their own solutions to the potential threat to their authority and to possible village discord. In the Javanese villages, local and village governments felt compelled to respond to such threats. In sub-districts in Java, three village governments funded surveys and allocated funds for social assistance from village budgets. Village officials also employed personal, cultural, and religious approaches to reduce tension related to mis-targeting issues in their communities. Village governments also arranged an informal redistribution of assistance and suggested that villagers who received the cash transfer share it with eligible villagers who did not receive assistance (Katiman 2023, 337–343). Taken together, these practices show a reinterpretation of government poverty assessment at the local level and a reconciliation with local ideas of poverty. Simultaneously, they point to a pattern of dealing with and working around programme problems.

Triangulating these outcomes with informants revealed that the excluded poor included people who had fallen into poverty after the survey, those misidentified by the survey or econometric targeting, and young households and the poor who had never been listed in the BDT. Those mistakenly enlisted as recipients included modest households that had been identified as poor and low-income families who had recently become prosperous. Notably, the non-poor recipients tended to be modest (“simple”) households, “capable” but non-affluent by village standards. As one village head registered, the system provided conditional cash transfers “to many people worthy of help, but often not to those most in need” (Interview, Aceh Besar, May 2017). Despite the arguments of officials (see above), the study found that many of these tended to have personal ties to the village apparatus. For example, in the Aceh Besar village, where most poor households are landless sharecroppers and labourers, the mother of the TKSK official, who owned her paddy field received benefits. The study found a member of the village council listed. In the Aceh Utara village, a preacher in the Islamic school, several hamlet heads, and the family of the village secretary, modest but not poor households, received CCT benefits.

A return trip to Aceh in 2019 revealed that the programme had continued to expand. Using PBDT data, the programme had increased from 14% to 21% of those surveyed in our stratified random sample, including some of the newly married families who had been overlooked by the earlier data. Validation processes seemed to have improved the targeting: conditional cash transfers now reached 23% of those classified as poor and meeting PKH criteria. However, 44% of those receiving benefits were still considered non-poor by the community poverty-ranking exercise. In this, a path dependency is at work, as past decisions constrained later outcomes. As district officials noted, once the system registered a household as poor and deserving of assistance, their names tended to be locked in. District officials observed that village heads, wary of creating lifelong enemies, avoided signing forms indicating that a household should be excluded. The disparity with those left out persisted.

In sum, the centralised system continued to generate a very similar list of beneficiaries, even though the number of recipients increased, and village leaders had attempted to correct it. Village leaders became frustrated as they witnessed the same erroneous lists returning from the centre. While distancing themselves from the process, village leaders proved unwilling to exclude those they considered less deserving. A high level of error continued. Hence, the ways of knowing that seek to render poverty governable involved standardising the social, turning it into legible and administratively implementable formats that combine simplification with over-complex, centralised delivery mechanisms. This generates significant problems of unfair distribution and mis-targeting. Simultaneously, state agencies embark on never-ending processes of repair, using new or recontextualised poverty technologies, importing and applying more complex approaches to solve the problem of the mis-targeted and misattribution of funds. This elicited a local politics of redistribution and stimulated village level workarounds and adjustments that, over time, generated a highly uneven order of entitlements.

In/exclusion: Changing the Order of Entitlements

It is crucial to understand how CCT knowledge practices affect the entitlements of the poor, here defined as the “broader domain of well-being and advantage” (Watts and Bohle 1993, 117). Changes in entitlements work out in several ways. First, as well as excluding people (as discussed above), by making direct payments to millions of households living in poverty, the CCT approach generates new forms of inclusion. During fieldwork, one beneficiary noted that “this money is important, in the first place, to send my child to school. But the money is also used for household necessities, such as food and clothing … It is not enough, but it helps” (Interview, Imogiri, Bantul, October 2018). Beneficiaries changed their consumption and spending behaviour to some extent, expending more on food, eggs, vegetables, and chicken. As another beneficiary said, “I cannot imagine making ends meet without this support” (Interview, Imogiri, Bantul, April 2019). While this generates new forms of dependence, it makes poor households feel “safer” and more “secure.” Some informants said they had even gained self-confidence. One noted: “We dare to borrow now, and we have more courage to look people in the eye” (Interview, Imogiri, Bantul April 2019). Local shopkeepers acknowledged this, reporting that “they get self-esteem from PKH. They live better lifestyles. If villagers have PKH, they are better now” (Interview, Imogiri, Bantul, April 2019). Yet, as other research has found, while the PKH cash transfers boost weekly household expenditure, reduce food insecurity, smooth consumption, and soften the experience of poverty, they are insufficient for moving recipients into productive livelihoods (Nasrudin et al. 2020, 7).

That said, we see the role of the CCT knowledge regime affecting the implementation of programme conditionalities and working in an overly rigid manner to create new forms of middle-class paternalism (see Veit-Wilson 2009). Every three months, programme facilitators (known as pendamping PKH) are responsible for assessing whether people have met programme conditions, by confirming that beneficiaries have participated in the health check programme and that children have attended school. Some go beyond this duty and scrutinise bills, checking for alcohol and tobacco consumption and the purchase of luxury items. They also audit internet usage and even monitor the grades of the children. Although programme facilitators cannot add people to the beneficiary list, since 2019, they have been able to suggest that people be taken off the list because they no longer belong to the target group (for example, if children graduate or if beneficiaries die) or if people do not fulfil their obligations. Local officials derive power from a knowledge regime that gives them the capacity to delete people from the list.

Facilitators vary with respect to how they implement their tasks. Some PKH fieldworkers meet only the formal requirements associated with checking and collecting numbers from schools and health posts. A cohort of long-serving and highly officious field operatives tended to be strict and had more skills to complete their tasks effectively. This may be because the training for facilitators in the earlier years was more intensive, and education requirements were higher. Other facilitators have little knowledge of social work and welfare. They bring their own – often stereotypical – ideas about poor people to their new role as facilitators. The close surveillance of people, and the power of facilitators to remove people from the lists of beneficiaries, open space for personal interpretations of their role and new forms of clientelism. A closer observation shows that interpretations of poverty and the required lifestyle changes of the poor go hand in hand with the personal ideas that individual facilitators have about the causes of poverty and the “unhealthy” practices of the poor, such as their views around cleanliness, discipline, and organisation.

Recipients’ reactions to the dominance of these officials are mixed. While all appreciate the financial benefits of the programme, a handful of informants openly complained about the conditionalities, or expressed fear of exclusion. The majority explained that they appreciated the guidance and support of the facilitators and hoped to gain access to more benefits in the future. Popular facilitators actively provided advice, for example, helping to improve the living situation of beneficiaries by finding additional sources of income, including advice on starting a business. Facilitators are urged to graduate 10% of their beneficiaries each year. This help and advice often occurred during the monthly family development skills workshops, during which officials try to “transfer” knowledge, skills, and values. Others have become brokers of information and knowledge for political reasons (Berenschot 2018). They help fix beneficiaries’ technical problems concerning getting their cash transfers and provide advice on other issues, including access to credit programmes, health clinics (for example, arranging a hospital bed), or family counselling. As he had become popular, one official told us he would be elected to the local government if he ran as a candidate in the next election.

The role of the facilitator is complex, contentious, emotionally taxing, and stressful, and facilitators view the job as difficult. Each facilitator has responsibility for 200–300 beneficiaries. The programme pushes facilitators to visit all beneficiaries regularly, coordinate and run training programmes, and complete all administration on time. Problems associated with poverty are significant and complex. Many beneficiaries are older than the facilitators, often women, leading to age and gender-related social complexities. Decisions to cut allowances are complex and fraught. Facilitators face a conflict of interest when providing advice about removing the PKH allowance from beneficiaries. As the PKH programme recruited most facilitators from the area in which they lived, they are sometimes neighbours of beneficiaries or their relatives or friends in the community. Therefore, it is hard for facilitators to be strict regarding their obligations: “We are from the area; we are afraid that people will turn against us,” one informant said. “How can I advise the department to remove my neighbour from the programme?” another noted (Interview Imogiri, Bantul, October 2018). A high proportion of PKH officers quit their job within the first year (Interview with enumerator, Imogiri, Bantul, April 2019).

As facilitators and local officials confront the social problems that the knowledge system elicits, notably the fact that so many of those receiving benefits are not “deserving” by local criteria, they respond in ways that lead to stigmatisation. Local government officials now place “poor household” stickers on the front of PKH households, to shame undeserving households into withdrawing from the programme because they face the views of excluded neighbours regarding whether they are entitled to benefits or not. In this way, programme mis-targeting inadvertently increases the sense of socially constructed shame imposed on the poor, and thus contributes to processes that divide the poor from the non-poor, reduces self-esteem and social capital, and inhibits agency. Indeed, research suggests that welfare policies that shame or stigmatise recipients may be psychologically scarring (see Walker et al. 2013). On this note, critics of residual CCT systems have argued that they lead to segmentation, and create a distance between the poor and other citizens left out of social provisions even if they remain vulnerable (Fischer 2018, 233–238).

As a side effect of the new CCT politics of knowledge, PKH facilitators who are often from middle-class families, thus exert moral and social pressure on poorer villagers, as they attempt to improve the lives of people with low incomes according to middle-class values. They may advise local people how to behave “properly,” directing beneficiaries to be more disciplined, dress well, clean the house, and spend less on credit for their mobile phones. Village heads and other village officials speak at the family development skills workshops, where they repeat these messages. In this way, both facilitators and village officials go well beyond the original goal of health and educational conditionalities and exclude people for social and cultural reasons. Moreover, such measures can reduce the agency of poor people, itself a key element in the positive literature about CCT programmes.

Hence, the study identified several constitutive effects linked to CCT knowledge politics. First, we see CCT knowledge producing patterns of inclusion, and higher consumption among poor recipients, and gains in self-confidence and agency in ways that align with programme narratives. Second, CCT knowledge is productive of a disciplinary normalisation. CCT knowledge underpins disciplinary institutions designed to observe and record the activities of the poor, seeking to internalise a type of subjectivity within the poor, and inducing them to conform to programme norms. As shown above, CCT knowledge systems also involve surveillance of the poor, paternalism, and the production of new forms of stigmatisation. In these ways, they underpin the emergence of non-egalitarian and asymmetric forms of power that aim to establish an idealised normative concept of poverty, and provide ways of dealing with it, that, via conditionalities, reward, punish or even stigmatise those who conform or deviate from the ideal recipient. Third, and somewhat paradoxically, during processes of translation and repair, there are local adjustments and strategic behaviours that undermine these normalisation processes. Local officials face problems and conflicts of interest, or they adjust, applying values and interpretations that lead them to employ programme conditionalities unevenly. As adjustments, project translation, and repair processes occur at the field level, operations are shifted away from the programme logics and the strict implementation of conditionalities or idealised representations found in the public descriptions.

The Constitutive Effects of CCTS: Conclusions

How do the knowledge politics inherent in conditional CCTs generate specific constitutive effects? First, building on the work of other scholars, this article discussed how CCT knowledge logics and practices establish a particular way of understanding poverty, by setting out how poverty can be measured and understood and framing how it can be governed. Analysts have acknowledged that poverty is a multi-dimensional problem that, as a social phenomenon, is “polyvalent and rich in meaning” (Dahler-Larsen 2014, 976). Hence, community-based, qualitative understandings need to be included. However, as Vetterlein (2012, 53) argues, “Actionable knowledge is subject to targets, measurable indicators and thus quantification.” Econometric targeting fits the requirements of actionable knowledge, and provides for targeting, effectiveness and rationing of benefits within available budgets, and audit reporting. While measurement subdues a messy reality, capturing only some dimensions, this “economisation of the social,” as Vetterlein (2012, 50) expressed it, enables standardisation into legible, administratively convenient formats for actionable policy. Yet, the CCT knowledge system involves simplifications that have critical implications. This, together with untidy, relational processes of data collection and checking, means that the knowledge system has limited capacity to “read” the causes of poverty in local contexts or represent poverty quantifiably. Despite the best efforts of implementors, the knowledge system generates unfair and exclusionary welfare distribution.

The knowledge system generates both national-programme and local-village practices of repair. The “technical problem” of mis-targeting (exclusion and inclusion errors) incites a never-ending process of introducing new data collection and beneficiary selection methods to obtain more reliable data. This constant re-contextualisation of poverty methodologies borrowed from elsewhere entails the endless improvement and reworking of systems for collecting and checking data (see Freistein 2016). In village contexts, there is an informal and largely undocumented repair of CCT practices as village leaders and local officials seek to lessen the socio-political impact of mis-targeting and avoid the corresponding negative impact on their authority and village social cohesion.

The knowledge system naturalises the logics and practices of social assistance, inscribing certain norms and procedures regarding how social assistance should work, thereby seeking to shape the social actions of those working in this social field. Furthermore, it shapes the order of entitlements and the capacity of individuals to obtain assistance. As suggested by the evaluation literature, it generates new forms of social inclusion, helps recipients gain access to assistance, and thus enables them to grow in confidence and address key depredations (Katiman 2023; Kidd and Bailey-Athias 2019). However, it also creates new forms of exclusion.

The CCT knowledge system likewise affects patterns of enfranchisement, that is, the degree to which individuals and groups “can legitimately participate in the decisions of a given society about entitlement” (Appaduri 1984, 481). It defines a new strategic landscape that those seeking benefits need to navigate. For instance, poor excluded households desperate to be recognised as beneficiaries react strategically. Poverty as a classification within the knowledge system is a resource: being classified as poor brings rewards. People manipulate surveys, lobby, and use social networks to try to get on the list of beneficiaries. However, CCT processes remain opaque. This has a disempowering impact: poor households who find themselves left out have limited means of seeking redress and tend to stay excluded for years. In contrast, others who are better off may remain included. In many cases, those who were excluded bitterly contest their exclusion by increasing demands for inclusion – actions that lead to tension, dissatisfaction, jealousy, the breakdown of solidarity networks, and even unrest.

This sometimes has more subtle, undocumented implications. For instance, CCT knowledge requires monitoring and sanctioning, and CCT programmes impose conditionalities. In the process of translation and repair, they institutionalise paternalistic practices and produce stigmatisation. Such techniques may generate a disciplinary normalisation of developmental practices in state apparatuses if fully implemented. However, under the pressure of local realities and social relations, local officials make strategic adjustments that undermine these normalisation processes, and programme conditionalities are applied unevenly.

Here, the econometric logics that underly programme design and evaluation, even when hybridised with community deliberation and surveys, run against village structures and practices ingrained in culture, political and social expectations, and behaviour. Villagers in rural Indonesia exist within – albeit unequal – elaborate reciprocity and mutual dependence networks. Given their inability to invest in networks of mutual assistance, those who need assistance most – the poorest of the poor – receive the least from social networks. Access to mutual help and local forms of social security are highly differentiated: “the poorer you are and the more support you need, the less you are protected” (Nooteboom 2015, 293). At the same time, villagers relate to the state more as clients than citizens, and villagers usually attempt to use social relationships to access state programmes (see Berenschot 2018; von Benda-Beckmann, von Benda-Beckman, and Marks 2000). Consequently, state programmes tend to be accessible for those who can invest in maintaining critical social relations. Hence villagers try to use such relationships to win inclusion while, by doing this, causing exclusion for others. At the same time, village leaders face pressures to support the inclusion of political allies, clients, or powerful actors – a predisposition that makes it difficult for them to exclude households openly unless they wish to make bitter enemies. Exclusion from benefits is considered an entitlement failure and impacts social cohesion and a household’s sense of inclusion in the community. Excluded villagers may shirk participation in village meetings and mutual assistance (gotong royong) and bitterly resent village leaders (see McCarthy and Sumarto 2018). Moreover, the individualised nature of cash transfers may weaken existing social support arrangements and mutual help and protection arrangements.

In contrast to local ideas of mutual support, the CCT process is a residual system that aims to differentiate the eligible from the ineligible according to a techno-administrative standard and procedure that prioritises efficiency. This does not map onto local notions of fairness or concepts of entitlement, and village heads cannot necessarily provide clients with benefits. Hence, the CCT knowledge system contradicts the “preexisting contexts, path dependencies, and the informal norms that regulate daily activities” (Olivier de Sardan and Piccoli 2018a, 53). Villagers would prefer a broader, more inclusive approach to a highly targeted system that leaves many poor out, a system that reflects established socio-cultural expectations and moral claims. However, a more inclusive model requires resources that the state and its donors are unwilling or unable to provide, except in crises (such as the COVID-19 pandemic). A more inclusive system would mean lower payments unless the social assistance budget was expanded.

This article has argued that, as CCTs both encounter disparities and set in motion new inequalities, the programme/system provokes a contentious local politics of distribution. Paradoxically, it also depoliticises distributional policy questions. This works in the following way. Poverty programmes (where decisions may be taken in meetings over technical designs) require that questions of social values and goals and the making of normative and political decisions be considered when developing redistributive structures and processes to move society towards a more just social order (see Fischer 2018). The model provides techno-bureaucratic procedures for selecting beneficiaries and delivering cash payments; a social problem is converted into a technical implementation problem (importing the best practice model, getting it to work, fixing operational issues). A firewall of technicalities, where only the experts understand the metrics, measurements, and targeting systems, cloaks discussions of poverty programmes (see Li 2007; Merry 2011). CCTs offer a solution to the symptoms of poverty (giving money to those who are poor) without addressing the structural drivers of poverty (why they are poor). In this way existing institutional practices and the boundaries of development discourses are re-affirmed, “excluding destabilising viewpoints” (Lyall and Havice 2018, 1549). However, the country that pioneered CCTs, Mexico, offers a useful insight regarding the shortcomings of pursuing cash transfers without supportive pro-poor structural reforms. Although Mexico had the oldest CCT programme in Latin America before it was replaced, poverty rates remained practically unchanged after two decades of implementation (Papadopoulos and Leyer 2016).

The work that CCT policy knowledge involves becomes visible during the messy processes that occur as the model is translated downwards. While the local government needs to manage the programme and enforce its rules, elected district and village leaders (as political representatives) persuade people that they cannot be reasonably held responsible for unfair or mis-targeted social welfare benefits. Yet, while protests emerge at various points, the system does not lead to a more polarised politics. There are objections and disappointments, but the excluded do not mobilise effectively to gain redress. Village leaders avoid conflict by giving the excluded access to other programmes, asking the poor to share their benefits, carrying out meetings off-stage to avoid open conflict, and shifting blame to higher levels of government. The technical ways in which “problems” are framed and delivered work to disempower local people. With the method and formula being too complicated, poorly explained, or kept secret, those on the ground may not understand targeting methods. The technocratic approach and bureaucratic procedures foreclose conflict and impede alternative framings of the problem, even while its managerial, technocratic, expert-led approach seems the most rational.

In conclusion, below its technical carapace, the CCT system involves a politics of knowledge that includes political decisions about redistributional goals, structures, and processes. Moreover, this generates political negotiation, strategic adaptation, and relational processes riven with contradictions and socio-political tensions during application. This study has contrasted this with how CCT knowledge systems are represented in the grey evaluation and econometric literature (see above), with their neat measuring, indicators and econometric targeting, surveys, and community deliberations. It was noted that these knowledge systems help define the social realities of which they are a part. This politics of knowledge serves the ideological and material interests of specific actors and their projects and shapes specific field outcomes. While it may be difficult to anticipate these effects in advance, this knowledge politics has shaped realities on the ground.

Facing this reality, how might civil society actors and government agencies develop fairer and more effective ways of addressing the needs and interests of the poor? State planners can continue to try to improve the programme, for instance, by strengthening the implementation capacity of local agencies (Widyaningsih, Ruhmaniyati, and Toyamah 2022, 27). Yet, this study implies that moving to more effective programmes for providing the poor with what Ferguson (2015, 50–51, 165–190) identifies as a “rightful share,” that is, towards a more inclusive system that fits with indigenous social and moral expectations, will need much more – a shift in the politics of knowledge. This will require modifying entrenched hierarchies of expertise to include types of knowledge and methodologies that do not conform to predominant econometric ideas of classification or quantification. This further implies shifts in the political economy toward supporting redistributive social reforms and policies that better address social, economic, and gender inequalities and provide integrated investments in the productive capacities of the rural poor. While the development literature provides discussions of alternatives, this is beyond the scope of this article. The point made here is that the specific politics of knowledge at work here obscure alternative ways of knowing and, by doing so, also deprive policy discussions of a vision for other approaches.

During the COVID-19 pandemic, facing a broader economic crisis, Indonesian state planners reintroduced a non-conditional cash transfer system and other schemes to assist those excluded from the cash transfer system (Yuda, Damanik, and Nurhadi 2021). However, the targeting methodology discussed here remains in place. As Indonesia moves forward, it remains to be seen how such issues will be addressed. To what extent will there be a turn to a new politics of knowledge, placing vulnerability, precarity, and structural inequality more effectively at the centre of social assistance?

Acknowledgements

The authors are very grateful to Andrew McWilliam, Vania Budiantoto, Sarah Cook, Katiman, Sumarto Mulyadi, and Ward Berenschot for their comments, support, or assistance at various stages of this work, and to Rudy Purba for managing the data analysis. Thanks to the support of the Population and Policy Research Centre at Gajah Made University Yogyakarta (PSKK-UGM) and Saiful Mulyadi and colleagues at the International Centre for Aceh and Indian Ocean Studies (ICAIOS) at the Syiah Kuala University for hosting the research in Java and Aceh. We also wish to thank the district governments and villagers in Aceh Besar, Aceh Utara, and Bantul for their generous hospitality and a wide range of Indonesian policymakers and programme managers at national and regional levels of government, including village leaders and District field staff of the Ministry of Social Affairs for generously sharing their reflections.

Additional information

Funding

This project received the support of the Australian Research Council for the grant “Household Vulnerability, Food Security and the Politics of Social Protection in Indonesia” [DP140103828].