Policy Research Working Paper 10883 A Learning Agenda for Community-Driven Development Responding to Complex Contextual, Evaluation, and Inference Challenges Patrick Barron Patricia Fernandes Stephen Winkler Michael Woolcock Social Sustainability and Inclusion Global Practice & Development Research Group August 2024 Policy Research Working Paper 10883 Abstract Governments and nongovernmental organizations around in CDD’s impact across four broad types of outcomes— the world utilize Community-Driven Development cohesion, inclusion, resilience, and process legitimacy. (CDD) approaches to address complex and overlapping The paper concludes by outlining a set of priority research development challenges. Despite consistent evidence on questions that will advance learning on CDD and provides some impacts of CDD—especially improvements in basic guidance on the empirical approaches and tools required services—there is significant variation in most outcomes to answer these research questions. The proposed learn- and several unanswered questions. This paper argues that ing agenda focuses on understanding variations in project the central task to advance learning on CDD (and simi- design, implementation modalities, and context, arguing lar complex development interventions) is identifying the that increased knowledge in these domains will help to conditions under which it works and the design and imple- optimize the impacts of current CDD projects, inform the mentation choices that will make it most effective within design of new projects, and develop an understanding of a given context. The paper provides an overview of CDD, what project designs are most scalable in different contexts. background on the existing evidence, and identifies gaps This paper is a product of the Social Sustainability and Inclusion Global Practice and the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at pbarron@worldbank.org or swinkler2@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team A Learning Agenda for Community-Driven Development: Responding to Complex Contextual, Evaluation, and Inference Challenges 1 Patrick Barron Patricia Fernandes Stephen Winkler Michael Woolcock JEL: D02; D71; I31; O22 Keywords: community-driven development; evaluation; local governance; sustainable development; learning 1 All findings and interpretations in this paper are solely those of the authors and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. We are grateful for the comments and constructive feedback provided by: Daniel Balke, Sean Bradley, Benjamin Burckhart, Sun Young Chang, Robert Chase, Narae Choi, Jose Cuesta, Abderrahim Fraiji, Jana El-Horr, Scott Guggenheim, Tracy Hart, Marcus Holmlund, Sarah Keener, John Loeser, Robin Mearns, Rafi Najam, Kaori Oshima, Nicolas Perrin, Ashutosh Raina, Vijayendra Rao, Audrey Sacks, Cyrus Samii, Janmejay Singh, Ifeta Smajic, Victoria Stanley, Zoe Trohanis, Makiko Watanabe, Katherine Williams, and Robert Wrobel. We also thank Louise Cord, Richard Damania, and Nik Myint for their guidance and feedback on the paper. Enduring errors of fact or interpretation are solely ours. Email addresses for correspondence: pbarron@worldbank.org and swinkler2@worldbank.org. 1. Introduction: Framing development challenges Countries and communities face an array of complex development challenges. Some challenges are technically complex, but solutions are largely known. 2 Other complex development challenges, however, require multidimensional, context-specific (localized) solutions that are not as easily designed or implemented (Pritchett and Woolcock 2004). This is the case for much work aimed at promoting socially sustainable development.3 But promoting change in such areas, and via processes participants regard as legitimate, can be challenging. The technical solutions to exclusion, social divisions, and a lack of resilience are often not clear from the outset and will vary from context to context, as will the ways in which people believe change in these areas should be pursued. This is the case for addressing development challenges such as enhancing local justice or negotiating land disputes. Challenges such as poverty reduction also require multidimensional solutions, especially when the context involves issues such as a fragile social contract, ongoing conflict, and weak governance institutions. In such cases, it is often not clear what is driving the observed problems, how they interact with one another, which of them should be prioritized over others (and on what basis), and what the optimal response could or should be in any given context. These multidimensional challenges are particularly acute in, but are not limited to, areas affected by fragility, conflict, and/or violence (FCV). Governments increasingly use Community-Driven Development (CDD) as an approach to address overlapping development challenges in difficult environments. Early CDD projects focused largely on building local infrastructure and have evolved in recent years to support areas such as climate adaptation, disaster response, local economic development, ex-combatant reintegration, and post-conflict recovery. Such varied objectives inevitably mean that the design, scale, and implementation of CDD projects differ. However, a commonality across CDD projects is that communities have decision-making power on how development resources are identified and prioritized. CDD is a commonly used approach for numerous reasons, especially in FCV contexts. CDD has shown a relatively unique ability to resist corruption and deliver basic services (Samii 2023), likely because of its incorporation of local accountability and transparency mechanisms, and the relatively wide discretion it grants trusted frontline implementers to customize responses to community idiosyncrasies (Hakiman and Sheely, forthcoming; Gibson and Woolcock 2008). By and large, it has proven to be an effective mechanism for delivering infrastructure and other public goods cost-effectively. In countries with limited state capacity at the national and/or local levels, CDD project mechanisms, including community councils and networks of facilitators, have demonstrated that they can deliver a wide range of types of development assistance. As governments increasingly adopt the CDD approach, often with World Bank financing, a wider set of Community and Local Development (CLD) projects have emerged. For example, some of today’s CLD projects are integrated into national systems and include components on devolved financing, local economic 2 Inflation, for example, is a technically complex challenge; skilled policy professionals may spend entire careers trying to tame it, but their energies will primarily be spent trying to craft and enact an integrated mix of technical solutions, the effectiveness of which will be readily apparent in changes in prices and carefully calibrated indices. The underlying mechanisms connecting inflation and interest rates are relatively well understood, such that raising rates a mere quarter of a point can have vast consequential effects on both domestic and foreign economies. Over recent decades, the complex mix of technical policy instruments deployed by central bankers has been refined to the point that most middle- and high-income economies rarely experience wild swings in inflation. 3 Communities and societies are more socially sustainable when they are cohesive, resilient, where groups are not excluded, and where development is done in ways that are broadly deemed ‘process legitimate’ (Barron et al. 2023). 2 development, climate mitigation and adaptation, and inclusion. 4 Defined widely, CLD currently accounts for around 10 percent of the World Bank’s portfolio, with projects across a range of global practices. 5 Despite consistent evidence on some CDD impacts, there is significant variation in most outcomes and many unanswered questions – especially regarding how variation in project design, implementation modalities, and contextual idiosyncrasies affect CDD outcomes. CDD approaches have a strong track record of effectively improving basic services (Casey 2018) and in some cases also improving economic welfare and participation outcomes. However, because CDD is often used in response to complex multidimensional challenges, and because CDD projects enable local discretion, impact trajectories over time are often non-linear and non-uniform, and may yield welcome and unwelcome outcomes that are not measured at all (Woolcock 2019). This makes it more difficult to draw conclusions about the nature and extent of projects’ effectiveness. Similarly, the mechanisms that link elements of project design to different outcomes (positive and negative alike) are under-theorized, as are the ways in which projects interact with elements of ‘context’. 6 As a result, there have been few empirical tests of these mechanisms (Avdeenko and Gilligan 2015 is a notable exception). It also means that evaluation approaches used to assess more conventional development interventions may not help us understand the conditions under which CDD yields certain outcomes for certain people in certain contexts in response to certain problems at certain scales of operation with certain implementation partners. 7 The central task – for CDD and similarly complex interventions – is thus identifying the conditions under which it can work, and the design and implementation choices that will make it most effective within a given context. Building on the evidence base we currently have, the next step in evaluating and learning from CDD programs is to better understand the conditions under which they work and how well they perform in specific contexts, in order to optimize the effectiveness of existing projects and support the design, implementation, and scalability of subsequent ones. CDD projects’ very designs, which allow for high levels of discretion from communities and project staff such as facilitators, means that what resources are used for and the ways in which project processes are implemented will vary considerably– between countries, projects, and, indeed, between areas targeted by a project in a given country. The goals of CDD projects are typically multifold, aiming to combine the provision of ‘hard’ development assistance (building infrastructure, etc.) with efforts to build – or at least engage respectfully with – local institutions and promote norms related to participation, inclusion, and the peaceful mediation of differences. 4 As of June 2023, the World Bank was supporting 375 active CLD projects in 98 countries with approximately US$45.2 billion in World Bank funding and an additional US$11 billion of borrower or other donor co-financing. The portfolio includes 27 of the 37 countries (73 percent) on the FY23 list of Fragile and Conflict-Affected Situations. 5 As of June 2023, the largest Bank-funded CLD portfolio is in Agriculture (US$15.5 billion), followed by Urban, Resilience and Land (US$8.2 billion), and Social Sustainability and Inclusion (US$5.8 billion). These are followed by Water (US$4.3 billion), Environment (US$4.2 billion), Social Protection (US$4.1 billion), and other global practices (Digital Development; Education; Finance, Competitiveness and Innovation; Governance; Health and Nutrition; and Transport, totaling US$2.1 billion). 6 Variation in how different mechanisms drive outcomes in different contexts also means that ‘lessons learned’ from one particular response to a complex problem may travel poorly to another context or population, even within the same country (Vivalt 2020). 7 Complexity is not always an obstacle to using standard evaluation approaches such as RCTs. Many complex interventions, including CDD projects, have been effectively evaluated with RCTs. Our point is that standard evaluation approaches alone cannot answer all questions about CDD’s impacts across different contexts or design and implementation choices. 3 Building the institutions and norms associated with CDD is key for development because the provision of cash and economic assistance alone cannot address problems related to the under-provision of public goods and how services are regulated. There is strong positive evidence on the impacts of cash transfers – both unconditional and conditional – in helping people out of poverty. 8 While cash can support people’s wellbeing, both through its clear utility for individual recipients and as a fiscal stimulus, it cannot address all development needs. One gap is in the provision of the public goods needed for people and societies to flourish – physical infrastructure, provision of health and education, and so on. Many of the goods needed for prosperity will not be supplied by markets without state investment in public goods. A second gap is the institutions needed to regulate and govern life, for example in ways that provide for goods such as security and justice. In most cases, cash programs do not strengthen or create such local governing institutions. The logic of providing cash and of providing community-level decision making through CDD is similar: that people know best how to use resources in ways that improve their lives. In this sense, cash transfers and CDD can be seen as compliments rather than competitors. Moving forward, it is most useful to use evaluations of CDD projects to ask two sets of questions. First, what are the range of outcomes on which CDD projects can have positive impacts – beyond access to basic services, which has been well documented? Such outcomes relate not just to direct measures of poverty reduction (such as consumption) but also other measures that are known to have positive impacts on poverty reduction, sometimes over the relatively short-run (improvements in public infrastructure) but also over the longer-run (the building of governance institutions, social cohesion, and collective action capacities). Second, under what conditions do CDD projects tend to have positive impacts on such outcomes? This will relate to elements of projects’ designs and implementation. It may also relate to the types of contexts (e.g., FCV vs. non-FCV; urban vs. rural; high vs. low social diversity) in which projects are implemented. This learning agenda will require new ways of evaluating projects and of generating knowledge from comparative analyses of impact, both within and across projects. The existence of significant variation in outcomes across project areas requires a greater understanding of the ways in which projects interact with local contexts. 9 Understanding such variation requires a greater focus on the development and testing of theories on the causal mechanisms that link project interventions to different outcomes (in particular, when those outcomes are also shaped by other factors outside of the control or scope of the project). At times this may require (a) the use of different methodological approaches beyond what is often deemed ‘rigorous’ (Pritchett 2023), such as randomized control trials; (b) tracking outcome measures other than those that have typically been included in project results frameworks and evaluations; and (c) focusing on longer terms outcomes and the processes through which they are shaped. A key priority for the CDD learning agenda is to improve the use of approaches that are well-suited to generating additional evidence on CDD impacts, including on the variation of these impacts over time, between groups, and across project designs, implementation modalities, and social contexts in ways that lead to refinements in design, implementation, and scaling. The paper sets out ideas for what this learning agenda could look like. 8 Fiszbein and Schady (2009) provided the first comprehensive articulation of the positive effects of conditional cash transfers on poverty reduction, though precise estimates of their long-term impacts remain difficult to discern (Millán et al 2019). 9 Almost six decades ago, Albert Hirschman argued that we need to think about “project behavior” and impact as being “rooted in such structural characteristics [of projects] and in the interaction between those characteristics and society at large” (Hirschman 1967, p. 4). 4 The objectives of this paper are to provide background on existing CDD evidence, identify gaps in theory and findings, and outline a path forward to address unanswered questions through a diverse set of research approaches suited to CDD. The paper identifies a set of priority research questions related to CDD’s impacts and provides guidance on the types of methods/tools that are best suited to address the questions. By extension, it also identifies broader inferences (and ‘lessons’) that can be drawn from individual studies, and how such knowledge might inform key operational decisions pertaining to initiating, designing, refining, replicating, scaling – and potentially shutting down some – CDD initiatives. However, the extent to which the paper can provide specific lessons for operational design is limited by the nature of the existing research. The priority objective of this paper is not to provide detailed operational guidance, but rather to identify and articulate a learning agenda on how to build more operationally relevant evidence moving forward. Over the long term, the learning agenda outlined in this paper can help optimize the impacts of current CDD projects, inform the design of new projects, and develop an understanding of what project designs are most scalable in different contexts. The learning agenda outlined in this paper is based on a desk review of existing evidence and methods, and by input from CDD practitioners and evaluators. The review of existing evidence builds on other relevant reviews including Mansuri and Rao (2012), Wong and Guggenheim (2018), and Winkler and Linneman (2022). Our paper adds to this past work by (i) summarizing CDD’s impacts on social sustainability outcomes; (ii) identifying what remains unknown about CDD’s impacts, especially when it comes to context, design, and implementation; and (iii) outlining a set of methods and tools to address these remaining questions. Despite these contributions, the paper has at least two key limitations. First, the review of evidence focuses on CDD as defined in the paper and is not meant to be a systematic review of all the literature on CDD. Broadening the definition of CDD – for example, to all types of participatory development – or applying a more systematic review of the evidence may yield different conclusions. Our narrower approach ensures there is more consistency across the projects we review and focuses on the types of projects that are often requested and financed at scale by governments. Second, the paper does not compare CDD’s impacts to those of other development interventions. This comparison is not straightforward given the wide variation in CDD designs; however, future research should explore comparisons between specific elements of CDD projects and other development interventions. This paper instead focuses on how variation in context, design, and implementation within CDD affects outcomes. The paper has multiple audiences. For development practitioners, it aims to summarize evidence as well as what we know about the elements of project design, implementation, and context that shape outcomes. This can help inform the next generation of operational designs of CDD and CLD projects. For project teams and evaluators, it will set out a series of questions and ideas on methods that can be used to assess project efficacy. Beyond this, it seeks to contribute to broader debates on how we conceptualize and measure ‘impact’ for complex development interventions that allow for high levels of local discretion in project implementation. While the paper focuses on evidence from evaluations of CDD projects, the learning agenda it identifies will have relevance for the broader set of CLD interventions. To this end, the paper is structured as follows. Section 2 provides an overview of CDD definitions and introduces a framework for conceptualizing CDD within the broader concept of CLD. Section 3 reviews the existing evidence, with a focus on CDD’s impacts across four components of social sustainability – inclusion, resilience, cohesion, and process legitimacy. Section 4 outlines the priority research gaps on CDD and provides recommendations on empirical approaches and tools that can help fill these gaps. Section 5 concludes and acknowledges the challenges to advancing the learning agenda on CDD. 5 2. What is CDD and how has it evolved? Community-driven development (CDD) is an approach to local development that emphasizes community control over planning decisions and investment resources. CDD programs prioritize community participation and voice in the planning, implementation, and decision-making processes of development. This approach enables communities to work in partnership with national and local governments to identify and manage community-level investments. CDD has become a popular development intervention because of its approach toward empowering local decision‐making and its reputation for getting resources to communities efficiently and building local institutional capacity and accountability. By focusing on bottom-up processes, CDD projects have shown an ability to deliver even where state capacity is lacking, where data is scarce, and in the highly fluid environments that characterize contexts affected by fragility, conflict, and violence (FCV) (Samii 2023). That said, it is important to stress that the purpose of this exercise is not to ‘prove’ that CDD categorically “works” (or not); rather, it is to identify some of the key conditions shaping the ways and extent to which CDD has worked (and for whom), as a basis for informing discussions about the likelihood that scaling or replication efforts might succeed. Government and nongovernment organizations around the world have utilized CDD approaches to deliver basic services, improve critical infrastructure, distribute social protection, respond to crises, and build local institutions. This includes large and small projects in places like South Sudan, Afghanistan, the Philippines, Myanmar, Rwanda, and Uzbekistan. The diversity of projects around the world underscores its popularity and importance as a development tool. However, this diversity can also be a source of confusion because not all projects have similar designs and objectives which, in turn, can muddle the conversation about the effectiveness of CDD. It is therefore helpful to briefly outline what we mean when we say CDD. CDD’s conceptual origins are rooted in ideas about community self-governance, efficient service delivery, and participatory development. 10 Communities will almost always have more knowledge about their needs, and how they can be addressed, than central planners (Easterly 2006). Potentially, they also have greater incentives to use funds in efficient and non-corrupt ways. Where communities hold funds and decide on how to use them, there can be a “short route” to accountable service delivery (World Bank 2004). In particular, in countries where state capabilities are limited, providing resources directly to communities, and giving them responsibilities to manage them, can allow for service delivery and lessen the load on states where capacity is stretched thin (Andrews, Pritchett, and Woolcock 2017). The early literature on social capital (e.g., Coleman 1988; Putnam 1993) highlighted how local social and cultural institutions shaped patterns of cooperation and collective action and how this influenced development outcomes. In principle, CDD has the potential to strengthen such institutions and associational life (see further discussion on this below). The participation induced by CDD projects could also have the potential to help build the foundations for democratized development. Such participation is seen by many as a right (Mansuri and Rao 2012). 11 In short, as Wong and Guggenheim (2018) argue, community-driven development programs often rest on the idea that they: (i) are an efficient way to quickly improve basic development infrastructure; and (ii) are popular with communities because they engage poor people as subjects rather than objects of development. 10 See the essays in the festschrift for Michael Cernea (Koch-Wiesser and Guggenheim 2021) on how social theory shaped the beginnings and evolution of social development work in the World Bank, especially Guggenheim (2021) on CDD. 11 Article 25 of the International Covenant on Civil and Political Rights outlines the universal right of all citizens to take part in public affairs. 6 A broad definition of CDD refers to any initiative that ‘induces’ community participation in the development process. Whereas ‘organic’ participation originates with civic groups, often in response to a government or market failure, ‘induced’ participation originates with a policy or development intervention (Mansuri and Rao 2012). Such interventions recognize the role of civil society in development and therefore attempt to mobilize civil society participation to advance development objectives. Participation here can mean many things, including voting on the selection of local representatives, engaging in meetings on the selection or design of local development projects, and contributing cash or physical labor to construct a public good or support a private enterprise group. Historically, this broad approach to CDD encompasses interventions such as social funds, microcredit groups, and community- based efforts to manage local services. 12 A more common definition of CDD refers to a narrower set of interventions that: (i) direct a financial resource transfer to administratively defined communities; and (ii) establish a formal mechanism for communities to maintain influence over this resource. These projects still incorporate some measure of induced participation, but the type of participation is more clearly defined and the resource transfer occurs at the community level. This definition distinguishes CDD from social funds, where decisions about community proposals are made by higher level authorities, and from cash transfer programs, where the money is directed to individual households. Instead, this traditional CDD approach often includes two core components: (i) allocating block grants to communities to be used for small-scale infrastructure such as roads, electricity or water supply, or improvements to schools and health facilities; and (ii) providing facilitation support to establish CDD committees with decision making power over the selection, implementation, management, and monitoring of projects financed through the block grant. Even within this narrower approach to CDD, there is significant variation in project design. For example: the size of the block grant can vary, which then impacts the types of projects that are selected; the number of cycles of block grants can vary (i.e. one year up to four-plus years of repeating grants); there may or may not be quotas for the participation of marginalized groups in decision-making fora or committees; lists of eligible activities for which resources can be used will differ including, in some cases, whether funds can be used for private goods; the level of facilitation varies, and in some cases facilitation will be provided by government staff while in others by non-governmental organizations; links between community identification of priorities and projects and local government planning process may be present or absent; and so on. As such, each instantiation of CDD is a unique combination of many and varied context-specific components, each of which is likely to entail the making of highly discretionary and transaction-intensive decisions by both implementors and participants. In recent years, CDD approaches have evolved to meet new challenges and become more programmatic to address systemic challenges. While many projects still follow a CDD model, with communities in control of decisions over the use of resources, others have adopted other community and local development approaches. These projects differ in two main ways from classical CDD projects. First, there is variation in the level of community decision making, with some projects providing local governments with control over the resources with input (but not final decision-making power) from communities. Second, there is variation in the location of the financial transfer, with some projects having a greater emphasis on linking communities to the devolved funding and planning processes of local governments. This can help build 12 It is beyond the scope of this paper to review this broader concept of participation. For more on how participation is shaped by opportunity structures, agency, and empowerment, see Alsop, Bertelsen, and Holland (2006). 7 local government capacity and accountability and tie community-identified initiatives into broader country systems and strategies. Figure 1: Conceptualizing CDD within CLD Figure 1 provides a framework for conceptualizing CDD and CLD along these dimensions. Community and Local Development (CLD) is an umbrella concept covering a wide range of projects focused on community- level development and CDD is one defined approach within CLD. CDD projects meet the two conditions outlined above: financial transfers are made at the community level and communities decide how to allocate this money. 13 Other CLD approaches may adopt just one of these two approaches. For example, some projects transfer funds to the community level but allow a local government body to decide how the funds are spent (type 1). This approach might be helpful in contexts where local government institutions are viewed as legitimate and where there is a push to strengthen government institutions and coordinate preferences across multiple communities to align with a regional development strategy. Other projects might keep decision-making power at the community level and transfer funds through local government (type 2) or keep both the decision-making power and funds at the local government level and simply allow communities to participate and provide inputs (type 3). Type 2 approaches may be useful where community-level capacities for financial management are weak. Type 3 approaches may be 13 Most CDD projects still involve local governments in some capacity. This may include local governments committing to finance operations and maintenance of CDD infrastructure and services (e.g., hiring teachers for kindergarten subprojects), providing oversight on decision-making over investments to ensure that the investments are feasible, incorporating CDD priorities into their own development plans, and co-financing CDD infrastructure and services. 8 pursued where the primary goal is to build local government systems, although it should be noted that many of the advantages of involving communities (see below) can be lost with this approach, in particular if avenues for participation are not substantive. This framework builds on previous attempts to summarize the menu of sub-national institutional arrangements in CDD/CLD projects (see, for example, de Regt, Majumdar, and Singh 2013). 14 Across these two dimensions, CDD and other CLD projects have both expanded to address evolving development challenges (e.g., climate mitigation/adaptation), to support private goods (e.g., local economic development) in addition to public goods, and to target certain populations in the community (e.g., women or displaced persons). This plethora of approaches and objectives makes it even more difficult to ascertain the impacts of CLD projects at the aggregate level. For this reason, this paper focuses on evidence and evidence gaps around CDD projects only, which share more uniformity, although many of the findings will have relevance for assessing and learning from other CLD projects. 3. Background on existing evidence and motivation for a new learning agenda Because CDD projects address complex and overlapping development challenges, they often have multiple objectives that vary across projects and contexts. This is an important difference from other development solutions where the intervention is limited to a single sector (i.e., education, transportation) and the expected outcomes are clear and consistent across contexts. Instead, CDD projects are multisectoral by default (because subproject selection is demand-driven) and typically aim to achieve a diverse set of objectives tied to both the transfer of funds and the process used to administer those funds. For example, a typical CDD project’s stated objectives might include: (i) increased access to basic infrastructure and services (via multisectoral and community-selected block grants); and (ii) increased participation in local development (via facilitation and institutions set up to administer the block grants). Beyond these stated objectives, CDD interventions are often expected to impact outcomes such as income, collective action, and trust. Of course, the extent to which these are reasonable expected outcomes depends on project design and context. For example, project designs that enable communities to implement road improvements 15 or to open pre-kindergarten centers might be more likely to increase income by unlocking access to markets and labor force participation. Projects that incentivize or require different groups within communities to work together to prioritize, select, and implement projects might be more likely to improve social cohesion through intra-community interactions. 16 14 A common critique of CDD is that it creates parallel governance structures that undermine existing local government units (LGUs). Figure 1 illustrates how sub-national institutional arrangements exist along a spectrum. Decisions about which arrangement to use often depend on the strength of existing LGUs and other contextual factors. For example, countries with traditions and cultural norms that favor bottom-up planning may be more conducive to the traditional CDD approach while those with stronger decentralized governance systems may be more conducive to a different CLD approach. Others have argued that as LGU ‘capacity’ and ‘responsiveness’ increases, CDD can be implemented through LGUs instead of through parallel structures. 15 Projects with larger block grants and those that allow neighboring communities to pool resources are more likely to see a larger selection of road subprojects. Similarly, large block grants are required to build or refurbish pre- kindergarten centers. 16 A common misperception about CDD is that it should contribute to social cohesion simply because it includes participatory processes that bring individuals together. However, this participation alone is unlikely to bolster cohesion, especially in communities that are socially homogenous or that have strong baseline levels of intra- community cohesion. 9 CDD’s multiple objectives have the potential to contribute to social sustainability, which the World Bank recognizes as critical to ending extreme poverty and promoting shared prosperity on a livable planet. The World Bank’s recently released framework on social sustainability outlines how tackling the development challenges of today requires a focus on promoting social cohesion, inclusion, and resilience – three core components of social sustainability (Barron et al. 2023). Social cohesion facilitates the trust and collective action necessary for development and negates the risks of divisions, in some cases violent, that can derail development. Inclusive development helps ensure that all people, regardless of identity or circumstance, benefit from development. Resilient societies and communities can better withstand shocks (man-made, economic, natural) without significant impacts on well-being. The framework also highlights the importance of process legitimacy – the ‘how’ of policy making, program design, and implementation, and the extent to which these elements accord with the values, norms, and priorities of people in a given context. 17 In principle, carefully designed and implemented CDD projects have the potential to combine formal administrative rules with local norms of deliberation to create the procedural legitimacy needed to make hard decisions, such as allocating finite funds to certain community groups and not others (Gibson and Woolcock 2008). Table 1: CDD outcomes for social sustainability Social sustainability component CDD outcomes Inclusion • Quality of and access to infrastructure/services • Participation in local development (especially among vulnerable groups) • Targeting of vulnerable regions, households, individuals Resilience • Economic welfare (income, assets, labor force participation) • Sustainable local institutions • Climate/hazard adaptation and mitigation capacity Cohesion • Shared purpose and trust within/across groups or in local government • Incidents of conflict/violence Process legitimacy • Perceived community influence over decision-making • Perceived fairness/transparency of local development processes Table 1 illustrates the connection between some common desired outcomes of CDD and the components of social sustainability – inclusion, resilience, cohesion, and process legitimacy. As emphasized above, CDD’s impacts on these outcomes will vary based on conditions such as project design, implementation, and the contexts in which projects are implemented. 18 One goal of this paper is to identify which factors determine whether CDD strengthens or undermines these social sustainability components or what additional research is required to identify these factors. 17 Process legitimacy refers to the extent to which a community or society accepts who has authority, what goals are formulated, and how policies and projects get implemented. It also encompasses the approaches for reconciling disagreements or tensions, especially among those who stand to bear the greatest costs. Process legitimacy shapes how stakeholders experience the development process and the extent to which they accept its outcomes. 18 The categories are not mutually exclusive as some CDD outcomes contribute to multiple social sustainability components. 10 The remainder of this section summarizes existing evidence on CDD’s impacts on inclusion, resilience, cohesion, and process legitimacy outcomes. The purpose of this review is to motivate a forward-looking learning agenda by highlighting the main unanswered questions related to CDD. The review draws on and extends other recent reviews of the CDD literature, mainly Wong and Guggenheim (2018) and Winkler and Linneman (2022). Our review contributes to this past work by summarizing CDD’s impacts on social sustainability outcomes as a way of illustrating how CDD projects can be a tool to address multiple complex and overlapping development challenges. We also contribute a focus on identifying what remains unknown about CDD’s impacts across core outcomes so that we can recommend approaches to answer these unknowns (Section 4). Our starting point for the review is the same sample used in Winkler and Linneman (2022), which focused on 104 publicly available impact evaluations, academic studies, project audit reports, monitoring reports, and internal evaluations of World Bank-financed CDD projects. 19 We also include prominent studies of non-World Bank-financed projects if they are particularly relevant to the CDD literature or to the identified gaps in the literature. As noted above, CDD is used to describe many different types of projects. 20 We apply the above scope conditions to ensure some consistency in the projects we are reviewing and to focus on the types of projects that are often requested and financed at scale by governments. 3.1 Inclusion Infrastructure and services One of the most consistent findings on CDD is that block grants to communities can improve access to, and the quality of, basic infrastructure and services. A meta-analysis of rigorous CDD evaluations, including World Bank-financed projects in Afghanistan and Sierra Leone, found that CDD projects can contribute to a robust and statistically significant 21 increase in the bundle of local public goods including access to clean water, hours of electricity, education and health outcomes, and infrastructure quality of education and health facilities (Casey 2018). A randomized control trial (RCT) of Afghanistan’s National Solidarity Program (NSP) found that women were especially likely to benefit from increased access to education, health care, and counseling services, and that the program significantly increased access to clean, protected water sources and increased usage of electricity (Beath et al. 2013). In Sierra Leone, the GoBifo project led to significant improvements in local public infrastructure, including more functional latrines, community centers, agricultural drying floors, and foot paths (Casey, Glennerster, and Miguel 2013). A follow-up study conducted 11 years after the start of GoBifo found that these positive impacts on functional public infrastructure persisted over the long run (Casey et al. 2023). A separate study found that GoBifo was less effective at providing local public goods in larger communities (Anderson and Magruder 2017). In the Lao People’s Democratic Republic, the Poverty Reduction Fund II (PRFII) project also increased access to protected water sources, improved community members’ perceptions of school quality, and decreased travel times to neighboring villages (World Bank 2016b). In the Solomon Islands, 19 Here, CDD projects are defined as those that include a direct financial resource transfer to administratively defined communities that the community can use for productive investments (i.e., small-scale infrastructure, revolving loans). 20 For a more detailed overview of the different types of CDD projects, see Wong and Guggenheim (2018) and Mansuri and Rao (2012). 21 The treatment effect for all CDD projects in the meta-analysis was 0.119 standard deviation units with a standard error of 0.025. 11 the Rural Development Program (RDP) improved satisfaction levels for access to roads, markets, water, sanitation, and electricity compared to baseline levels (Neelim and Vecci 2013). Variation in the size of the block grant, the number of cycles of block grants, and the types of sub-projects can affect how CDD impacts infrastructure and services. For example, the average block grant size for projects with consistent positive impacts (in Sierra Leone, the Philippines, and Indonesia) ranged from US$ 11 to US$27 per capita. All communities in these projects also received two or more rounds of block grants. In comparison, projects with block grant sizes around US$ 1 to US$ 5 per capita and only one round of block grants – as was the case for a non-World Bank project in the Democratic Republic of Congo – produced weak or null results on infrastructure and service outcomes (Mvukiyehe and van der Windt 2020). The size of the block grants shapes the types of subproject that get implemented because small grants preclude communities from using the project funds for more costly projects such as the rehabilitation of roads or electricity provision. Of course, the idea behind CDD is that communities get to decide what project will have the most meaningful impact in their lives, but project design also plays a role in shaping the options on the table for communities. In rare cases, research shows that CDD projects can also improve health and education outcomes. For example, the National Program for Community Empowerment - Generation (PNPM - Generasi) 22 program in Indonesia created significant positive impacts across a range of health and education outcomes, including a 10 percent decrease in childhood malnutrition (Olken, Onishi, and Wong 2011). This decline in childhood malnutrition was detected 15-18 months after project implementation but did not persist 27 to 30 months after implementation, in part because childhood malnutrition declined dramatically in both treatment and control areas between baseline and endline measurement. In Cambodia, the Commune and Sangkat Fund (CSF) project reduced infant mortality in treated villages by 3.2 percent compared to the mean baseline value (BenYishay et al. 2019). In Senegal, the National Rural Infrastructure Project (PNIR) significantly improved the nutritional status of children even in PNIR-eligible areas without completed subprojects, potentially because of spillover effects (Arcand and Bassole 2008). In Nepal, the Poverty Alleviation Fund (PAF) project increased school enrollment among 6- to 15-year-olds by 14 percentage points, including girls and disadvantaged caste/ethnic groups (Parajuli et al. 2012). However, most evaluations of CDD projects have either not measured human development outcomes or found no impacts on immediate human development outcomes. An evaluation of the National Human Development Initiative (INDH) program in Morocco found that small positive economic impacts did not translate into any improvements in early childhood development outcomes (El-Kogali et al. 2016). This is not necessarily surprising given that education-related investments were only a portion of the project’s investments and education outcomes were not part of the project’s development objective indicators. 23 The project also included investments in electrification, rural roads, and drinking water, all of which produced high rates of return (El-Kogali et al. 2016). Participation Evidence suggests that CDD projects can contribute to increased participation in local development, including among women and vulnerable groups. CDD’s investment in facilitation and the establishment 22 PNPM-Generasi is a hybrid CDD cash transfer program. 23 The INDH evaluation acknowledges several program design features of the Indonesia PNPM program that help explain why that project did impact health outcomes, including incentivizing and supporting communities to select projects that would improve health and education indicators. 12 of local councils requires community participation. Often, projects also incorporate features that require participation among women and other vulnerable groups. In Afghanistan, NSP mandated gender parity in community councils leading to a significant and long-lasting increase in the number of women participating in village development activities (Beath et al. 2013). In India, the Pudhu Vaazhvu Project (PVP) project nearly doubled the number of women who attend the Grama Sabha, and increased their frequency of speaking by nearly 45 percent (Parathasarathy et al. 2017). However, it remains unclear when and why this participation spills over to non-CDD community development activities and how much participation is necessary to achieve such outcomes. Several studies find that communities with CDD projects experience an increase in a range of non-CDD community activities such as attendance at village meetings. 24 However, less is known about why this spillover occurs in some contexts and not others, or how this spillover affects other community development outcomes (see also the discussion below on Social Cohesion). A critique of CDD is that it demands time from vulnerable community members who could be spending that time on other more productive tasks. However, research does not support this argument and instead shows that community members value the opportunity to participate (see the process legitimacy section below for a discussion of this research). At the same time, future research should attempt to understand, in different contexts and scales of operation, how much participation is necessary and which modes of participation are best to achieve effective subproject implementation. In other words, more research is needed on the facilitation investments of CDD projects. Targeting While many CDD projects invest in public goods and services that benefit all community members, evidence shows that these projects tend to effectively benefit the poorest areas of a country and vulnerable members of a community. For example, an evaluation of the PNPM project in Indonesia found that households in the poorest sub-districts were more likely to experience positive welfare impacts (Voss 2012). Separate projects in Indonesia also effectively benefitted areas with conflict victims (Barron et al. 2009) and areas with low baseline health and education indicators (Olken, Onishi, and Wong 2011). Within communities, evaluations have found that CDD projects benefit the poorest households (Voss 2012; Labonne 2013; Nkonya et al. 2012; Deininger and Liu 2009) and women (Beath et al. 2013). CDD’s ability to target the most vulnerable in a community has been attributed to its deployment of participatory methods (Parajuli et al. 2012). The benefits to vulnerable households can vary by country and project. For example, although the Community-Based Reintegration Assistance for Conflict Victims (BRA-KDP) in Indonesia effectively targeted conflict regions, within villages it did not benefit conflict victims any more than non-conflict victims (Barron et al. 2009). In a Burkina Faso project co-financed by the International Fund for Agricultural Development (IFAD), IFAD found that the project lacked a targeting strategy and monitoring system 24 In Lao PDR, the PRFII project significantly increased attendance at general village meetings, including non-PRFII meetings, overall and for poor women (World Bank 2016b). In the Philippines, after the first cycle of KALAHI-CIDSS subprojects, household participation in village assemblies increased by 20 percentage points (Labonne and Chase 2008). In India, households in NRLM villages reported participating in at least three community meetings in the last year, compared to an average of fewer than one meeting in non-NRLM villages (Institute of Rural Management Anand 2017). Similarly, almost one-third of households in the Tamil Nadu Empowerment and Poverty Alleviation Project (PVP) areas reported attending the last Grama Sabha meeting, 31 percent higher than non-project areas (Khanna et al. 2015). 13 capable of capturing information on vulnerability and that this allowed existing village governance norms to prevail and limit the benefits to vulnerable groups (IFAD 2020). As CDD projects evolve, and some target specific marginalized groups, there are additional questions about CDD’s impacts on inclusion outcomes that should be explored. This includes questions about whether activities layered onto the CDD model can effectively foster positive social norms, such as norms regarding gender equality, intimate partner violence, or forced labor. More research is also required on how variation in the amount and type of facilitation in CDD projects affects norms and behaviors regarding the role of marginalized groups’ participation in the CDD project and other aspects of public life. Table 2: Inclusion – Summary of the evidence What we know: • CDD is an effective tool to improve quality and access to basic services and infrastructure • CDD increases participation in local development processes What is unknown: • How variation in the type and quality of facilitation affects outcomes • When and why participation spillover occurs and how this spillover affects other development outcomes • How much and what type of participation is necessary to achieve efficient subproject implementation • Why some projects are more/less effective at reaching vulnerable regions and households • The extent to which CDD projects foster pro-social norms (i.e., regarding women, displaced persons) • What implementation modalities, structure, and phasing work best to increase participation in local decision making 3.2 Resilience Economic welfare (income, assets, markets, labor force participation) Traditional CDD projects rarely include an explicit objective to increase income or private assets; however, evaluations show that the investments in small-scale infrastructure and services sometimes lead to modest improvements in economic welfare. A meta-analysis of CDD projects, including World Bank- financed projects in Afghanistan and Sierra Leone, finds that, overall, projects tend to have a positive and statistically significant effect on economic welfare measures such as household income, consumption, assets, and employment (Casey 2018). 25 For example, in Sierra Leone, the GoBifo project contributed to an increase in household asset ownership, a 30 percent increase in the number of petty traders, and a 13 percent increase in goods available for local sale within treatment villages. In Cambodia, the CSF project reduced poverty in treated villages and increased daily consumption spending (by 7 percent) and monthly 25 A review of the International Fund for Agricultural Development (IFAD) CDD projects (many of which are not World Bank-financed) conducted by their Independent Office of Evaluation (IOE) also finds that CDD projects often improve household consumption and living standards (IFAD 2020). These economic welfare effects are most likely driven by CDD’s investments in infrastructure and services. However, research also suggests that CDD’s approach of forming groups that increase cooperative activities and interactions with local leaders can also contribute to increased economic returns (Blattman et al. 2016; Macours and Vakis 2014). 14 income (by 13 percent) for households in these villages (Boret et al. 2021). CSF’s effects on poverty reduction persisted over the nine-year study period and the projects also significantly increased economic development in the surrounding area 26 (BenYishay et al. 2019). In Nigeria, the Second National Fadama Development Project (Fadama II) increased the mean income of beneficiaries by 40-60 percent and increased the value of productive assets owned by community groups, especially for the poor and for women’s groups (Nkonya et al. 2012). In Nepal, an impact evaluation found that the PAF project increased per capita consumption by 19 percent (Parajuli et al. 2012). Participation of villages in the BRA-KDP project in Aceh, Indonesia, which primarily provided cash assistance to conflict victims rather than public infrastructure, was associated with an 11-point reduction in the share classified as poor by village heads. Livelihoods-focused CDD projects that use a self-help group (SHG) model to mobilize poor women have also improved household income and savings. 27 Evaluations of the flagship India National Rural Livelihoods Mission (NRLM) have generally found that the program improved household incomes (Institute of Rural Management 2017; Kochar et al. 2020; Singh and Pandey 2019). For example, a rigorous impact evaluation found that an additional 2.5 years of NRLM SHG membership increased total household income by approximately 19 percent (Kochar 2020). 28 Studies of NRLM, JEEViKA, and the Cambodia Livelihoods Enhancement and Associations Among the Poor (LEAP) project found that these projects increased household savings, consistent with the SHG’s model of encouraging and facilitating savings practices (Pandey et al. 2019; Kochar et al. 2019; Datta et al. 2015; Ban et al. 2020). Findings are less conclusive on the impacts on household consumption, expenditures, and assets. In India, an impact evaluation found that the Andhra Pradesh District Poverty Initiatives Project (APDPIP) increased per capita consumption by 11 percentage points; however, studies of the NRLM, Tamil Nadu PVP, and JEEViKa projects did not find these projects to improve households’ total or consumption expenditure (Institute of Rural Management Anand 2017; Khanna et al. 2015; Datta et al. 2015). Relatedly, while there is some evidence that these projects increase assets for beneficiary households, this effect varies depending on the asset type (Institute of Rural Management Anand 2017; Khanna et al. 2015; Datta 2015). Other studies show more mixed or null results on economic welfare. In Afghanistan, an RCT of the NSP found that the project had strong positive effects on perceived economic welfare, especially among women, but no strong effects on household income, consumption, or assets (Beath et al. 2013). Similarly, an evaluation of KALAHI-CIDSS showed no effects on household poverty status or labor force participation and earnings (Beatty et al. 2018). Variation in a project’s overall design, the type of subprojects funded, and the per capita allocation of funds likely affect CDD’s impacts on economic welfare outcomes. 29 For example, subprojects chosen by the community that consist primarily of improving clean water and sanitation or health services may not necessarily have direct impacts on per capita income in the short or 26 As measured by nighttime light output. Each completed CSF project increased the NTL output by 20 percent. The median area received four CSF projects and experienced a gain of nearly 80 percent in NTL output. Rural transport projects are the main driver of these results and the impacts are especially large in more densely populated rural areas (BenYishay et al. 2019). 27 CDD projects with a SHG model are similar to traditional CDD projects in that they provide transfers of funds to the community level and communities decide how to allocate these funds. However, in the SHG model, the funds are typically transferred to a community-based savings organization and the funds are used for livelihoods activities rather than for infrastructure/services. 28 A recent review of evidence from SHG programs in the South Asia region, including non-World Bank-supported programs, reported mixed findings on the impacts on income (Javed et al. 2022). 29 For example, early versions of a non-World Bank financed project in the Democratic Republic of Congo allocated only ~US$ 1 per capita so it is not surprising that evaluations found no evidence of positive impacts on household economic welfare (Humphreys et al. 2012). 15 medium-term. Increases in per capita household expenditure in Senegal were limited to villages with agriculture and education subprojects, not water or health (Arcand and Bassole 2008). Morocco’s INDH created small positive economic impacts in the initial years of the project, but these effects dissipated over time (five years after baseline data collection) (El-Kogali et al. 2016). More research is needed on a new generation of CDD projects that include explicit activities and investments on livelihoods. 30 These projects vary in design but often include a component that provides grants for private goods to individuals or small groups within a community. The grants might be accompanied by access to information (e.g., value chain assessments) and training (e.g., business development and financial literacy) and often include a focus on women’s economic empowerment (WEE). A mixed-method mid-term review of the Kenya DRDIP project, which includes a component that provides capacity building and financial support to community livelihood groups, points to positive impacts on access to capital, contributions to household income, and women’s savings (World Bank 2023). However, a lack of livelihood diversification and time poverty limits women’s engagement and returns from the investments. In Aceh, Indonesia, a community-based forest protection program included livelihood grants as a form of in-kind compensation to forest rangers and to promote more environmentally sustainable economic activities. The project led to significant positive impacts on all subjective measures of economic welfare and a reduction in self-reported illegal logging activity (Paler et al 2015). However, there was little impact on actual behavioral change related to forest usage and an unexpected increase in mining activity. Overall, more research is needed on the livelihoods focused CDD projects, including on how project design can address perverse incentives and barriers to the realization of benefits among women and other marginalized groups. As CDD projects evolve to include explicit livelihood activities, evaluations should also explore economic welfare outcomes that are rarely investigated in the existing literature. This includes measurement of labor force participation and impacts on the formal and informal economy. Institutions There is some evidence that the community institutions established through CDD projects are durable and adaptable tools to administer services and respond to shocks. For example, a recent study from Afghanistan found that institutions created under the National Solidarity Project (NSP) and Citizen’s Charter Afghanistan Project (CCAP) endured in a context where other governance institutions had collapsed (Samii 2023). Moreover, the NSP/CCAP community development councils (CDCs) were adaptable to new service delivery needs that go beyond their original purpose of delivering infrastructure. More specifically, the CDCs were used to manage primary schools handed over from an INGO. The study found that administering education provision through the CDCs was more cost-efficient and just as effective (measured through children’s learning, household satisfaction, and cost) as continued NGO administration. 31 This finding counters the common critique that CDD institutions (and large investments in facilitation) are not meaningful or sustainable outside the scope or longevity of the project. Similarly, in Sierra Leone, community councils remained operational a decade after their creation (Casey et al. 30 This new generation of CDD projects combine elements of the traditional CDD approach with the SHG model. They often include a component focused on public infrastructure and services along with a component focused on livelihoods at the community level. 31 Currently, the World Bank and eight UN agencies are using the 35,000 CDCs to deliver and coordinate humanitarian and basic services such as food aid, cash transfers, water and sanitation services, and livelihoods support at the local level, ensuring that women and the most vulnerable receive assistance. (Source: Correspondence with Susan Wong and Robert Wrobel.) 16 2023). At the same time, there is little evidence to suggest that these investments create spillover effects on other local institutions. For example, the same Sierra Leone project did not generate meaningful impacts on collective action, decision-making processes, or the involvement of marginalized groups in local affairs beyond the scope of the project (Casey et al. 2012). A more recent study from Afghanistan shows that the CDCs in Afghanistan were used to facilitate cash transfers to vulnerable women in Afghanistan following the Taliban takeover in August 2021 (Callen et al. 2023). Climate adaptation/mitigation and responses to other shocks CDD projects have frequently been used to quickly channel funds to communities following a shock. Wong and Guggenheim (2018) document a range of examples including using local CDD platforms to rebuild over 100,000 houses in Aceh, Indonesia, following the Indian Ocean Tsunami of 2004, responses to a 2005 earthquake and 2010 floods in Pakistan, and responses to Typhoon Yolanda in 2013 in the Philippines. The presence of facilitators and village committees, as well as mechanisms to quickly pass resources to the local level, allowed for timely responses to emergency needs. CDD platforms have also been used for other purposes. In Aceh, for example, following a peace agreement, information on the content of the accord, and common messages agreed by the former rebel group and the Indonesian government, were distributed through the CDD project structure and information on post-conflict needs was collected by project facilitators (Barron and Burke 2008). However, there is a lack of data documenting the short- and long-term impacts of CDD on various outcomes of resilience at the community level. This is despite arguments that CDD is a particularly strong, scalable solution to promote climate and disaster resilience (Arnold et al. 2014). Additional research is needed to understand if and how CDD interventions integrate climate risks into local planning capture local knowledge on climate, promote resilience, and increase climate adaptation and mitigation co- benefits. Table 3: Resilience – Summary of the evidence What we know: • Investments in small-scale infrastructure can improve household and community economic welfare • CDD institutions can be durable and adapted for other uses What is unknown: • What types and size of projects are most likely to create economic welfare impacts • The impacts of CDD projects on outcomes such as labor force participation, formal and informal economies (especially among newer livelihood-focused projects) • What factors make a CDD institution durable over the long term. If CDD institutions are used for activities beyond the project, why is this the case? How does variation in local governance moderate this? • What CDD interventions work best to: i) integrate climate risks into local planning; ii) capture local knowledge on climate; iii) promote resilience; and iv) increase climate adaptation and mitigation co-benefits? • How does bundling facilitation with cash transfers (at community or household level) affect economic welfare, public infrastructure, and social sustainability outcomes? • Do CDD institutions help communities respond to natural hazards and other shocks? How? Why? For how long? 17 3.3 Cohesion Shared purpose and trust Existing evaluations find that CDD has minimal or null effects on measured social cohesion outcomes. 32 Rigorous evaluations of projects in Afghanistan, Sierra Leone, Nepal, and Lao PDR find that those projects have no substantive effect on a range of social capital and social cohesion outcomes (Beath et al. 2013; Casey et al. 2012; Parajuli et al. 2012; World Bank 2008). 33 Similarly, Casey’s (2018) meta-analysis of CDD projects, including those in Afghanistan, Sierra Leone, and Sudan, finds that, overall, the projects do not improve trust outcomes. 34 In Kyrgyz Republic, a CDD project had moderately positive impacts on a shared sense of unity between social and ethnic groups but no strong impacts on trust in other people or in local government (Esenaliev et al. 2018). An analysis of the KALAHI-CIDSS program in the Philippines finds that the project increased trust in strangers (Labonne and Chase 2008). In Thailand, the Social Investment Fund (SIF) project positively affected several social capital outcomes including a greater sense of self-sacrifice for the common good and improved vertical connections to formal authorities (Chase et al. 2010). However, the same Philippines and Thailand projects, along with a project in Indonesia, were associated with decreased trust in neighbors in some contexts, perhaps because the introduction of new funds increased tensions (Barron, Diprose, Woolcock 2011; Chase et al. 2010; Labonne and Chase 2008).35 An analysis of The Gambia’s Community Development Project (CDP) finds that households in treatment villages reported fewer social links and less participation in community-based organizations (Heß et al. 2021). In contrast, a large mixed methods study of KDP in Indonesia found improvements in relations across a range of identity cleavages in project villages compared to matched control sites (Barron, Diprose, and Woolcock 2011). Although many hypothesize that CDD projects should improve social cohesion, there are several potential explanations for these mixed and null effects. First, most CDD programs do not include any specific social cohesion activities and are not designed with any explicit purpose to improve social cohesion (Chatterjee, Gassier, and Myint 2023; Avdeenko and Gilligan 2015). Second, CDD communities are often quite homogenous and have high baseline levels of trust, cohesion, social ties, and civic engagement (Casey 2018; Samii 2023). 36 For example, evaluations in Sierra Leone and Afghanistan found high baseline levels of willingness to trust another community member with financial transactions (95 percent and 85 percent of respondents, respectively) and membership in at least one social group (75 percent in Sierra Leone) (Casey et al. 2013, Beath et al. 2013). Third, evaluators use varying definitions of social cohesion and may not be using the most context-appropriate tools to measure these concepts. Fourth, in many fragile, 32 See Chatterjee, Gassier, and Myint (2022) for a discussion on diagnosing gaps in social cohesion and for a review of evidence on social cohesion interventions in development projects. 33 A review of IFAD’s CDD projects also concludes that there is limited evidence to confirm a causal link between CDD and social capital. However, the report states that the participatory planning and capacity-building approach improved the extent to which rural people and their communities meaningfully participated in making local development decisions (IFAD 2020). 34 An RCT of a non-World Bank financed CDD project in Liberia found that the project reduced social tension and improved individuals’ trust in community leadership (Fearon et al. 2009). Individuals living in CDD communities were also 7.5 percent more likely than those in control villages to cooperate with their neighbors to solve community problems. 35 In Indonesia, the CDD approach effectively resolved any increased tensions (Barron et al. 2011). 36 It is possible that communities with low intra-community trust are less likely to be targeted by CDD projects or that measures of trust are not picking up intra-community conflict. 18 conflict, and violent environments where CDD projects operate, issues of security, rule of law, and justice are oftentimes higher-level concerns for communities that may eclipse immediate service delivery needs. These issues are beyond the scope of CDD and other interventions (World Bank 2011). Lastly, it may take years if not decades to change social norms and build trust in some societies and project evaluation periods of three to five years may not be sufficient time to detect these changes (Woolcock 2019). Social change itself may be highly non-linear and non-uniform, following widely variable impact trajectories over time, space, and scales of operation – implying that empirical impact evaluation claims from all types of complex interventions need to benchmark against a corresponding theory of change articulating what it is realistic to expect by when. As such, relatively small-scale, short-term, intra- community interventions may be fated to yield modest positive (but most likely “mixed”) results in social domains such as norms, networks, and trust. That said, the attainment of any positive development outcomes in FCV contexts such as Afghanistan, Sierra Leone and Sudan are surely a commendable achievement; moreover, if we take seriously the ethical imperative to regard participation as a right, then perhaps it is more helpful to regard prevailing forms of social norms and networks as a potentially vital mediator, rather than an expansive goal, of ‘induced’ development projects. Vertical trust and the social contract Research that looks specifically at CDD’s effects on the citizen-state relationship shows that CDD projects can improve citizens’ confidence in the government. An analysis of the Afghanistan NSP, one of the largest CDD projects in the world, finds that CDD strengthens legitimacy in the state because it delivers development projects that are responsive to community needs (Parks et al. 2019, Olken 2010). More specifically, the study finds that CDD projects that are responsive to community needs increase public confidence in local and district government and increase the likelihood that villagers are willing to use government courts to resolve disputes. 37 Respondents living in CDD districts in Afghanistan have also reported slightly better perceptions of the national government’s performance than those living in non- CDD districts (The Asia Foundation 2018). In the Philippines, trust towards local officials increased by 11 percentage points in villages that received funding from the KALAHI-CIDSS CDD project (Labonne and Chase 2008). However, households in these villages also requested fewer services from the local government and see fewer non-KALAHI-CIDSS projects implemented in their village, suggesting that the CDD project may substitute for other local investments. 38 Another study of KALAHI-CIDSS finds that it improved knowledge and awareness of local officials and governing bodies and that development funds (including non-KALAHI-CIDSS funds) were more closely aligned with preferences of residents (Beatty et al. 2018). Yet the number of CDD projects measuring impacts on the social contract is still limited. Conflict and violence CDD, at least in its earlier years, was sometimes seen as a mechanism for bolstering local conflict mediation skills and addressing some of the drivers of localized conflicts (Barron et al 2011). There has been relatively little research on whether it can have such impacts. Studies in Afghanistan, Sierra Leone, 37 These results are especially strong when the baseline relationship between citizens and the state is low. As citizens develop more confidence in the state, the strength of this effect diminishes. 38 A review of IFAD’s CDD projects finds that, “CDD projects contributed to governments’ decentralization efforts where they were embedded in established and sustainable local government structures. CDD projects operating outside government structures had little to no direct impact on local governance” (IFAD 2020, pp viii). The report also found mixed evidence on the sustainability of rural institutions that were created or strengthened through CDD. 19 and the Philippines found mixed impacts of CDD on conflict (Beath, Christia, and Enikolopov 2018 Casey, Glennerster, and Miguel 2013; Crost, Felter, and Johnston 2014). One study (Arcand, Bah, and Labonne 2011) found that the CDD project in the Philippines was associated with a decrease in violent events involving the rebel group, the MILF. While we would not expect CDD projects to have significant effects on lowering levels of violent conflict, it would be fruitful to explore further whether CDD can be effective in managing the conflicts that can be generated by the introduction of resources into fragile areas. An assessment of KDP in Indonesia found that local conflicts related to the project were less likely to turn violent than those related to other development projects but that there was little evidence that the project had a positive impact on conflict at the aggregate level (Barron, Diprose, and Woolcock 2011). Table 4: Cohesion – Summary of the evidence What we know: • Many CDD communities have high levels of existing intra-community trust • CDD projects are unlikely to affect macro levels of conflict but may be effective at addressing development related conflicts at the local level What is unknown: • Under what contexts is CDD more likely to impact cohesion? What types of cohesion? • Does variation in implementation modality moderate CDD projects’ impacts on views towards the state? • How can the alignment of incentives for local government and communities be strengthened for longer-term service delivery? • What core design adjustments are necessary in CDD approaches to create impacts on social cohesion (facilitation + timeline for community/type of subprojects and size of grants)? • Does incorporating narrative strategies (e.g., perspective-taking) into CDD facilitation work help to reduce prejudice towards marginalized groups? If so, how can this approach be scaled across projects? • What are the long-term payoffs to investments in community social capital? • Are existing forms of community-level social capital best understood as a means to attaining development outcomes or is ‘enhancing social capital’ a realistic development outcome in its own right?) 3.4 Process Legitimacy Perceived influence over local decision-making One way that CDD projects may affect process legitimacy is by shifting households’ perceived influence over local decision-making. For example, in Lao PDR, participants in PRFII communities were significantly more likely than those in control communities to say that their community had influence over village decision-making and that the local government sought community inputs in the village (World Bank 2016b). In Afghanistan, 77 percent of surveyed respondents reported a person like them could influence the community development council (CDC) decision-making (The Asia Foundation 2018). In a rare case in the Philippines, the KALAHI-CIDSS project decreased individuals’ belief that they have agency to improve the situation in their community despite increasing overall participation in community organizations and knowledge about local government (Beatty et al. 2018). Data suggest that this result may be driven by 20 individuals who do not feel confident participating in community development activities, perhaps because participation is new to them. 39 Fairness and transparency The participatory process may help reduce elite capture and yield subprojects that better reflect community preferences. In the Afghanistan NSP program, secret ballot referenda were found to reduce the influence of elite male preferences on the prioritization of subprojects and move subproject locations further from the village headman’s house (Beath et al. 2017). In Senegal, an evaluation found that, although elite capture may be present, 40 it is mitigated by the increased number of women involved in the decision-making process (Arcand and Bassole 2008). Qualitative evidence from the Indonesia PNPM Urban program reflected that community respondents attributed purportedly low corruption levels to community participation and monitoring (World Bank 2013). In another test of elite capture, Casey et al. (2012) found that when villages in Sierra Leone were given an asset and told they could use it for private or public use, nearly every village used the asset for public purpose. Evidence from the KALAHI-CIDSS program in the Philippines shows that projects reflect local preferences overall and are not aligned only with the preferences of wealthy and educated households (Labonne and Chase 2008; Casey 2018). In Malawi, a recent assessment of what participants across nine communities in a CDD project themselves think of CDD found that participants understood and appreciated its distinctive role (when carefully facilitated) in accommodating their preferences and values (Krishna et al, 2024). Table 5: Process legitimacy – Summary of the evidence What we know: • CDD process can increase communities’ perceived influence over local decision making • CDD projects can improve confidence in government What is unknown: • How different approaches to decentralization or other implementation modalities impact outcomes on process legitimacy? • How can CDD improve the quality of interactions between local government and citizens and the types of procedures required to reassure citizens the process is fair? • How do CDD projects affect citizen’s perceptions of local government performance and credibility? How does variation in CDD project’s linkages to local institutions moderate these impacts? • Are there additional activities that might help improve trust and confidence in government? 39 This is another instance where empirical and policy inferences drawn from data taken at face value, no matter whether a given ‘impact’ is deemed positive or negative, may be misleading; more precisely understanding what such findings mean requires interpreting data in the light of theory, experience, and contextual realities. Requiring ‘participation’ in a development project, for example, in contexts where participants have little experience with public deliberation and/or where entrenched elites may regard such initiatives as a direct threat – and where initial participatory efforts do indeed leave participants feeling humiliated and subjected to hostile pushback – are unlikely to yield responses from participants indicating that they now feel ‘empowered’ to positively influence outcomes in their community. That such a reality may well prevail in many contexts suggests that even empirical findings of ‘no impact’ could in fact be a significant achievement. 40 Here, elite capture is a measure of whether a village received a completed project. 21 4. Outlining an operationally relevant research agenda In this section we outline the priority empirical and theoretical gaps to focus on to advance learning on CDD and provide guidance on the empirical approaches and tools required to fill these theoretical gaps. We structure this section around the: (i) outcomes that remain underexplored; (ii) sources of variation in CDD impacts; and (iii) methods and tools that can be used to address the gaps in evidence. The discussion is largely focused on evidence and learning related to traditional CDD projects rather than other forms of CLD projects, which often include a focus on other outcomes beyond public goods (such as livelihoods). The section does not provide evidence on the comparative efficacy of CDD vis-à-vis other CLD approaches. Nevertheless, the learning agenda outlined will allow for future comparisons between CDD and other CLD approaches, which can help inform when different approaches are most appropriate. 4.1 Outcomes There is fairly robust evidence on many of the key outcomes that CDD projects have traditionally sought to achieve, related to economic sustainability. A panoply of evaluations has shown CDD, with some exceptions, to be a useful mechanism for delivering relatively small-scale public goods (such as local roads, irrigation systems) in cost effective ways that are popular with the public. In particular, in areas where local state institutions have limited capacity, reach, or are unpopular, the CDD mechanism, where funds are transferred to communities who hold decision-making power over their use, has by and large proven to be effective. There is less, and sometimes conflicting, evidence on the extent to which improvements in access to services through the provision of such infrastructure leads to improvement in measures of economic well-being such as consumption. (Newer CDD projects which include components supporting local economic development appear to be more effective here.) There are more gaps in the evidence base on some outcomes related to social sustainability. As the discussion in Section 3 highlighted, while there is reasonably solid evidence on most inclusion outcomes, there has been relatively little research on the cohesion, process legitimacy, and resilience outcomes. Evidence on the effectiveness of CDD programs to extend social networks and promote pro-social norms is mixed at best, though the complex nature of issues being addressed, the contexts in which they reside, and CDD interventions themselves suggests that research findings on these effects are bound to be ‘mixed’ long into the future. There are a few potential reasons for these gaps. First, it may be that most current and past CDD projects are not designed in ways that would likely influence these outcomes. In terms of aspirations to improve cohesion or reduce conflict, it is often unclear what the theory of change is on how projects could affect these. For example, research has shown that in many rural communities where CDD projects are implemented, cohesion within villages is already strong. If this is the case, then it is not obvious how more meetings involving more people within villages would enhance what is already robust levels of cohesion. Indeed, in many areas, cohesion issues are more marked between different communities, yet most projects do not include mechanisms to bring people from different villages together. Similarly, while local factors are important in shaping the likelihood of localities being affected by large-scale conflict, often forces much larger than community cohesion or resilience shape the incidence and spread of conflict (Kalyvas 2006). Similarly, it is unclear how most CDD project designs would affect perceptions of government performance beyond satisfaction with the project, especially as the interface of many projects with beneficiary populations are facilitators who are often not seen as representatives of the 22 state. Using CDD to strengthen the social compact may require additional project activities such as public information campaigns (Baseler et al 2023). Second, CDD’s focus on outcomes beyond inclusion (particularly beyond indicators related to public infrastructure and services) is relatively new and evaluations have not caught up to the new focus on resilience, cohesion, and process legitimacy outcomes. An increasing number of CDD operations are seeking to influence outcomes such as trust in government and between communities, and resilience to climate change and other shocks. Yet attempts to measure whether CDD projects are contributing to such outcomes are still in their infancy. Third, research designs that are typically seen as ‘gold standard’ may not be best suited to measuring these outcomes. In part, this is because of the inherent challenges associated with the gap between concepts and measures for many social phenomena; in part, it is because surveys, which collect data with a relatively short interface between the researcher/enumerator and the interviewee, are often not the best way to collect data on sensitive social issues. It is also because the timelines through which change happens in areas such as institution-building and social norms change often far exceed the lifetime of projects (and hence the periods for which their impacts are evaluated). A null finding in the short run does not necessarily mean a project had no impact; it may have laid the foundation for future changes. Fourth, and related, these outcomes typically do not make it into project results frameworks. This, in part, is because these outcomes are typically not fully under the control of projects but are a result of the ways in which projects interrelate with a host of exogenous forces, many of which are non-linear and that may take years to be realized. Incentives for project team leaders are to select indicators which projects have a strong chance of achieving. Project documents often imply that projects will shape these outcomes (e.g., “CDD can help reduce fragility”) but teams, understandably, do not want to be held accountable for achieving these aims. But these imperatives have worked against our understanding of how external actors can support processes such as institution building which are key for development. Moving forward, it is important to unpack these broad concepts and to map project designs to a theory of change with clearly defined outcomes. Project documents often implicitly claim they will affect social outcomes but do not measure these outcomes and are not always designed in ways that would logically allow them to be achieved. There is a need to be realistic about what projects can achieve and to map interventions more clearly to outcomes. In parallel, it is important for evaluators to pay attention to actual project investments and not to test performance against outcomes that projects have never been designed to achieve. By extension, it is also entirely plausible that unanticipated positive and negative outcomes may be associated a given CDD intervention, which qualitative approaches may be best suited to discerning (see Rao et al 2017). It is equally important to identify measurable indicators for these outcomes and to use appropriate methods to measure the indicators. Some of these lend themselves to quantitative measures, but some may not (and, in most cases, the use of multiple methods will be suitable to triangulate and better understand survey data). Not all outcomes will have simple indicators that can be included in project results frameworks. 23 4.2 Sources of Variation We know relatively little about what causes variation in the extent to which CDD projects achieve their desired outcomes. We do know that design factors such as the size of the block grant and the number of block grant cycles influence the types of subprojects that are selected and that this, in turn, creates variation in outcomes such as economic welfare and access to basic services (including improvements in human development outcomes and in the quality of interactions with local governments). As is true for all complex intervention, implementation quality and contextual compatibility are also likely to be key factors (see below) (Woolcock 2022). However, there have been very few research designs that allow us to tease out systematically how and why such variation occurs. 41 As noted earlier (Section 4.1), there is a need for greater clarity in aligning project design with theories of change that clearly outline projects’ intended outcomes in terms of specific aspects of inclusion, resilience, cohesion, and process legitimacy. Projects can choose multiple design and implementation routes to reach a similar objective and, therefore, we can expect significant variation among operations of a particular “sub-type” – i.e., those that aim to improve village infrastructure and therefore access to basic services or those aiming to improve consumption or income. A research agenda that would allow us to assess how, under what circumstances, and for whom “CDD works” will need to explore how variation in design and in the quality of implementation can lead to very different outcomes (within a specific sub-set of operations with the same overall objectives so that we are comparing like with like). Three sets of factors may affect variation in impacts. First, the impact of design choices beyond poverty targeting of investments, size of block grants, and number of project cycles needs to be better understood – particularly as it relates to social inclusion, cohesion, resilience, and process legitimacy. One such source of design variation is the extent to which CDD programs choose to align with local government planning and budgeting processes. This can be in terms of the volume of resources available for programing at the subnational level, the extent to which local governments play a role in the selection and co-funding of such investments, the composition of the local institutions responsible for implementing and overseeing subproject activities, the role of budget transparency measures put in place by local governments, or the extent to which they leverage local government performance grants or incentives. Design choices in this area would be expected to significantly influence process legitimacy and resilience outcomes – yet they remain under-analyzed. Another source of variation is the type and role of local institutions in CDD projects. There are significant differences in the way in which local institutions (village committees, project specific committees) are selected, strengthened through training, given decision-making autonomy, required to enforce program rules, and held accountable for their decisions. All of these factors (and others) may lead to very different outcomes in terms of the effectiveness and sustainability of these institutions. Facilitation is critical in this regard. Sources of differential impacts may include: (i) the ratio of facilitators to the size of the community; (ii) skills and experience requirements for facilitators (including language skills in remote and conflict-affected areas, level of expertise for dealing with more complex aspects of 41 A notable exception is Ensminger (2017), which used a comparative review of the KDP project in Indonesia and the ALP project in Kenya to identify key differences in project design that shaped the nature and level of corruption in each project. 24 facilitation and planning – e.g., elite capture, stigma against vulnerable and marginalized groups); (iii), gender, age, and ethnic background – which can play an important role for particular aspects of facilitation, such as promoting gender equality; (iv) whether the facilitators are a cadre hired, trained, and managed by government institutions or whether the facilitation role is a specialized function outsourced to an NGO (or UN agency, which is often the case in FCV); and (v) the supervision structure in place for facilitators to deal with challenging aspects of their work, and coaching/mentoring arrangements to uphold team morale. How these choices influence quality of implementation is key to answering the question of when and how CDD programs can deliver on intended results. 42 Considerably more research is needed in this area. There are other operational choices which may affect how project designs shape the likelihood of achieving desired outcomes. Another important dimension of design likely to lead to significant variations in outcomes – including in terms of process legitimacy and trust in government – are issues related to the flow of funds, systems for effective community procurement (where relevant), and those for ensuring quality assurance (e.g., engineering support, quality control of investments and/or market assessments, technical trainings, and market linkages for CDD programs focusing on livelihood interventions). Design choices in these areas – especially the compatibility of government systems with project procedures that enable r funds to flow, the establishment of clear operational procedures for financial management and procurement, and the outsourcing of (or in-house capacity for) specialized technical roles – can significantly impact the credibility of programs. Secondly, fidelity to design during the implementation of projects is a likely source of variation in impacts. CDD projects present specific challenges when it comes to ensuring that processes laid out during the design phase are followed through – particularly when it comes to facilitation and the participatory process. This is particularly true for large-scale, national CDD platforms where there might important trade-offs between getting to scale and maintaining the initial intensity of the intervention as designed. CDD programs cover large areas with investments that are spread out and which employ a very large numbers of facilitators, hence relying to a significant extent on the training of community groups and volunteers. There is a level of design complexity required for programs that have the explicit goals of including marginalized and vulnerable groups in decision-making for example, or those that focus on building climate resilience by integrating available national/sub-national data on climate vulnerability and community knowledge of climate risks (which likely reside in very different epistemological spaces). To understand variation in CDD impacts (again within specific sub-types of projects aiming at the same type of result) it will be critical to understand the extent to which design choices have been translated into practice as intended yet still locally accepted. Expanding/standardizing the use of process evaluations in CDD programs is an important step in assessing fidelity to design and ensuring that implementing agencies can take timely corrective action. For example, a process evaluation of the Philippines NCDDP showed significant discrepancies between the participatory planning cycle as designed and as understood and implemented in practice by community facilitators. Assessing the effectiveness of quality assurance mechanisms put in place at the project level (supervision, community-based monitoring, training and coaching of facilitators, technical oversight of infrastructure investments) through qualitative analysis and process monitoring, and creating feedback loops with implementation, must be an important part of a future research agenda. This is needed so we can gain a better understanding of the causal transmission 42 See Bold et al (2018) for a formal assessment of the different outcomes attained in an identical education program (itself previously determined to “work” in India as a result of an RCT) implemented in Kenya by two different organizations (the government and an NGO). 25 mechanisms between the design and implementation phases, and how these affect the extent to which projects work (or do not work). A third source of variation, and where more research is needed, is to understand how contextual differences shape outcomes. One issue is how projects adapt to different situations within the same country. Conditions for implementing CDD projects will be different in areas affected by conflict or violence than in peaceful areas. Impacts may vary when projects are implemented in areas where there are likely to be more marked social divides than in homogenous communities. Impacts may vary in more urbanized and more rural areas. 43 Some communities may have deeper ingrained traditions of collective decision-making which CDD can build on, while in others power and decision-making may be less diffused. Some of these factors will be known at design stage and often larger scale/more mature CDD programs will take some of these variations into account and set out specific implementation modalities for conflict- affected areas (e.g., Myanmar or Indonesia), disaster prone areas (the Philippines), remote areas with the presence of indigenous groups or ethnic minorities (the Philippines, Indonesia, and others) for example. In spite of these significant adaptations to context, there has been limited analytical work to track variations in impacts across different types of areas targeted (against the primary project objective and intended results). Analysis focusing on context has tended to be more process oriented – to understand the extent to which design and implementation modalities are fit for purpose in these different contexts. Less attention has been paid to varying impacts. Only one largely qualitative study, an assessment of KDP on local conflict in Indonesia, has looked at variations in impacts across targeted villages (Barron, Diprose, and Woolcock 2011). The lack of such studies is problematic because we would expect the achievement of outcomes to differ greatly across countries and projects and across areas targeted by the same project. This will be an important aspect to take into account in a forward-looking research agenda – particularly to help us understand how CDD can be one in a range of institutional instruments that can be effectively deployed where there are significant risks of re-direction of funds away from community needs or to sustainably allocate climate financing to communities. Table 6 outlines a set of priority research questions to fill evidence gaps on social sustainability outcomes across the three main sources of variation – design, implementation, and context. 43 There is a particular gap in evidence on CDD in urban settings. This gap is important, since the nature of social life in ‘settled’ rural communities is highly likely to be different in degree and kind in more transitory and diverse urban communities. Whether and how CDD interventions can adjust to these latter circumstances is vastly underexplored. 26 Table 6: Priority research questions by outcome and source of variation Source of Variation Design Implementation Context Inclusion • How much and what type of participation is • What implementation modalities, structure, • Under what conditions are there spillover necessary to achieve efficient subproject and phasing work best to increase benefits from participation in subproject implementation? participation in local decision making? implementation processes to non-CDD • Can CDD projects be used to foster positive • What approaches to processes? social norms, especially regarding women, hiring/training/supervising facilitators are • How do CDD programs learn from non-CDD displaced persons, and marginalized groups? If adequate/effective to increase participation processes in the local community? so, what design elements (including potentially among excluded groups (their influence over additional activities) are required for this? decision making)? • What are the main barriers (structural or behavioral) that marginalized groups face in participating effectively in CDD programs? How can these barriers be overcome to ensure more inclusive decision-making? Outcome Category Resilience • How does bundling facilitation with cash • How does variation in implementation • Do CDD institutions help communities transfers (at the community or household level) modality, structure, or phasing affect respond to natural hazards and other affect economic welfare, public infrastructure, impacts on climate-related outcomes? shocks? How? Why? For how long? and social sustainability outcomes? • If CDD institutions are used for activities • What CDD interventions work best to: i) beyond the project, why is this the case? integrate climate risks into local planning; ii) How does variation in local governance capture local knowledge on climate; iii) promote moderate this? resilience; and iv) increase climate adaptation • Do parallel local non-CDD community and mitigation co-benefits? councils exist? If so, how do they interact with the CDD council during project implementation, and how does this affect the durability of CDD councils in the post- CDD period? Cohesion • What core design adjustments are necessary in • How does variation in implementation • In what type of contexts can CDD CDD approaches to create impacts on social modality moderate CDD project’s impacts on approaches enhance social cohesion? cohesion (facilitation + timeline for views towards the state? Which types of social cohesion (linking, community/type of subprojects and size of • How can the alignment of incentives for bridging, and bonding) and what elements grants)? local government and communities be of design are important for these? 27 • Does incorporating narrative strategies (e.g., strengthened for longer-term service • In what contexts are spillover effects in perspective-taking) into CDD facilitation work delivery? terms of participation in local government help to reduce prejudice towards marginalized decision-making (including my marginalized groups? If so, how can this approach be scaled groups) more likely/more likely to be across projects? sustained? • What are the long-term payoffs to investments in community social capital? Process • How do CDD projects affect citizen’s perceptions • How does implementation modality affect • How do CDD projects affect citizen’s Legitimacy of local government performance and citizen’s views of the project? perceptions of local government credibility? How does variation in CDD project’s performance and credibility? How does linkages to local institutions moderate these variation in the baseline strength of local impacts? institutions moderate these impacts? • Are there additional activities that can help improve trust and confidence in government? • How can CDD improve the quality of interactions between local government and citizens and what procedures are required to reassure citizens the process is fair? Other/Cross- • To what extent do differences in approaches to • How does the sectoral focus or unit of cutting the training of community members community participation (e.g., farmers’ (community committees involved in cooperative vs. women’s livelihood group) prioritization/decision making) – curriculum, affect outcomes? sequencing, intensity, focus on community leadership skills) lead to more sustained outcomes? 28 4.3 Connecting Methods, Tools, and Theory Some of the evidence gaps and null findings are a result of weaknesses with the designs of evaluations, the inherent challenges of ‘measuring’ certain social variables, and the accompanying theories invoked to draw inferences from observed data. These weaknesses are often driven by research designs that do not consider the unique design and implementation features of CDD projects. For example, because investments are multisectoral and selected by communities, CDD evaluators should not necessarily rely on average effects over the entire treatment area (e.g., one should not expect to see health impacts at the aggregate level if 90 percent of communities selected road projects); similarly, null effects on average may mask wide variation in outcomes for different groups and contexts, from which much can be learned if an array of different research methods are deployed (see below). This matters because some external studies seek to measure the impacts of CDD projects on outcomes which the projects themselves were not designed to achieve. Similarly, CDD projects require skilled, experienced, and persistent facilitation, which by design grants high levels of discretion to implementers so that they can accommodate contextual idiosyncrasies in real time. Some facilitators will be extraordinarily good at this; others will be mediocre; while still others will, of necessity, be young, inexperienced, and no match for seasoned local leaders with vested interests in maintaining the status quo. It is thus neither possible nor desirable for facilitators to be performing uniformly, yet this is precisely the requirement of an experimental design. As Rao et al. (2017) report from a related project promoting citizenship training, a null impact on average was due in part to positive outcomes in well-facilitated communities and negative outcomes in poorly facilitated ones. For present purposes, it is important to note that such nuanced interrogations of a ‘null’ average impact were made possible by the explicit incorporation of qualitative research methods. For some harder-to-capture outcomes, such as ‘social cohesion’, these concepts have not always been measured well – potentially leading to null findings or erroneous positive or negative effects. Similarly, different multisectoral outcomes are likely to unfold along trajectories with varying timeframes and ‘shapes’; the economic impact of a new bridge, for example, may be almost immediate and readily measurable, while attaining greater gender equality or social inclusion for marginalized groups may take decades, with little to show for diligent efforts after (say) three years (Woolcock 2022). Taken at face value, these latter outcomes would be deemed ‘failures’ compared to the clear impact of the bridges, yet it is surely unreasonable to expect inherently different objectives to be achieved in the same timeframe. Moreover, policy objectives with historically ‘flat’ impact trajectories create the additional challenge of making it unclear whether “no impact” in the initial stages is a result of a flawed design and poor implementation, or a result to be expected after such a relatively short time from even the most carefully designed and diligently implemented project. Assessing empirical results in the light of a theory of change is crucial for all development interventions, but especially those whose impact trajectory is non-linear, non-uniform, highly variable over time, or influenced by neighboring programs. Child nutrition programs, for example, can yield rapid positive results – and thus were initially greeted with great enthusiasm – but their overall impacts can wane considerably over time if associated cash transfer programs increase prices for perishable foods, thereby leading – as they did in a study in the Philippines – to rising levels of stunting in the children of non-recipients (Filmer et al 2023). 44 44 See also Majumdar and Rao (2017), who found that the skilled facilitation – i.e., deft accommodation of local contextual realities, achieved via explicit ‘co-production’ with participants – characterizing the initial stages of a 29 Box 1 summarizes some common mistakes to avoid when evaluating CDD projects. Box 1: Common mistakes to avoid when evaluating CDD 1. Focusing on outcomes that are unrelated to a project’s investments or theory of change (ToC) Evaluations should focus on indicators in the project’s results framework and other outcomes identified in the project’s theory of change as having a clear connection to the project’s investments and/or design. For example, an evaluation should not focus on individual education outcomes (e.g., test scores) if the project does not include investments (e.g., teacher training) that would impact these outcomes. 2. Poor or imprecise measurement of concepts There are often many ways to define and measure concepts, such as social cohesion, that might be outcomes of CDD projects. Evaluators should take the time to understand how CDD projects will impact these concepts and across which aspects of the concept. For example, with social cohesion, evaluators need to identify which types of cohesion (bonding, bridging, linking) are likely to be impacted, through which mechanism (shared purpose, trust, etc.) and at what level (intracommunity, intercommunity). Too often, evaluations measure complex concepts (e.g., cohesion) with generic indicators (e.g., trust) that are not relevant to the project’s design. The very essence of such complex concepts, however, means that aggregation across and comparison between contexts – and thus broad claims about ‘impact’ – will be inherently problematic, no matter how ‘precise’ measures might be in any one instance. 3. Failing to incorporate the multi-sectoral and demand-driven nature of investments into the research design CDD projects, by design, create wide variation in the type and number of sectoral investments across communities. In one CDD project, 80% of communities might select road investments, 10% education, and 10% health. If the project provides four rounds of block grants, there may be further variation within communities – one community might select road investments for all four cycles while another selects three road investments and one education investment. Evaluators must account for this in the research design and ensure they are analyzing impacts for investments that actually occurred. For example, analyses should not focus on aggregate-level health impacts if only 10% of communities implemented a health project. 4. Failing to adapt the evaluation to changes in implementation plans or variable capability Several factors can cause changes in CDD implementation plans over the course of a project, including a sudden shift in the development challenges (e.g., a pandemic or displacement crisis), challenges and lessons learned from early phases of implementation, or a shift in government or donor priorities. Some evaluators resist adapting their research design to the updated implementation plans because doing so can compromise the internal validity of the research design. However, failing to update the research design can result in an analysis with unwarranted conclusions because it does not capture or reflect what factors enabled effective implementation on the ground. Evaluators should build flexibility into their research design and, when necessary, accept when a research design is no longer able to measure the impacts of a project that did not implement as originally planned. Existing research has not captured the long-term effects of CDD projects. Evaluations are expensive and a project’s timespan is often linked to election cycles; and as the years pass, attributing current outcomes to initiatives in the distant past becomes increasingly difficult. Even so, engaging in sensible debates about the ‘effectiveness’ of development interventions, as noted above, requires acknowledging that measurable shifts in certain domains (e.g., gender equality) may take decades, even centuries, and unfold along non-linear trajectories. This has important implications for CDD projects, especially for those striving successful CDD project focused on poverty reduction fell away markedly when the project rapidly scaled up, wherein facilitators “were mobilized quickly with a homogenous and fixed script that lacked the kind of improvisation that characterized the first phase, and which failed to include diverse stakeholder interests, objectives, and voices.” When combined with concomitant shifts in India’s political climate at the time, the adoption of uniform implementation protocols “significantly reduced the intensity of participation and its concomitant social impacts.” 30 to use such projects to build durable, credible, and impactful local institutions. Two rare examples of papers that look at long term impacts are Samii’s (2023) work on Afghanistan and Casey et al.’s (2023) paper on Sierra Leone’s GoBifo project. It should also be noted that social institutions in place at any particular level of development are likely to change as a result of the broader development process itself, most notably as communities become more integrated into broader networks of exchange, and as rising levels of education prompt graduates to explore more spatially diverse forms of employment (e.g., in cities rather than villages). Truly isolating a statistically significant ‘impact’ of a single development project on these social transformational imperatives is thus inherently very challenging. However, supporting long-term evaluations to generate more evidence-informed theories of change to guide policy inferences drawn from assessments conducted over short timeframes is largely beyond the patience, resources, and tools of contemporary social science. What evaluators can more fruitfully do is allocate more energy and time to understanding how, why, where, and for whom particular CDD interventions ‘work’ in the present, and less to engaging in ex-post meta-debates about whether, on average, the entire portfolio of CDD interventions ‘work’ (as provoked by conclusions drawn from many systematic reviews 45). For all the reasons outlined in this paper, the answer to the latter question is surely almost always going to be “the evidence is mixed”. It’s what inferences researchers and policy makers draw from such conclusions that matters most. Overcoming these weaknesses and filling the gaps in research will require a diverse set of tools that include and go beyond traditional evaluation approaches. It will also require greater recognition by those who ‘consume’ (as opposed to ‘produce’) research that learning from and accurately discerning the effectiveness of CDD interventions requires the deployment of methods optimally suited to the types of problems CDD addresses and the types of response mechanisms being used. The findings elicited by these particular methods, in turn, need to be interpreted correctly, along with their implications for policy and practice. All data must be interpreted in the light of experience and theory – a carefully articulated theory of change, for example – which specifies from the outset what outcomes it is reasonable to expect a given project to attain by when in a particular context. In short, CDD interventions are complex and multi- dimensional but this complexity alone is not an obstacle to rigorous evaluation and learning. Many complex interventions, including CDD, have been evaluated with a range of tools including RCTs, long- term studies, and qualitative analysis. We are simply calling for the full range of tools to be considered and appropriately selected based on the research question and theory of change. This should provide assurance that our proposed learning agenda is sound, aspirational, and feasible. CDD projects integrate constituent elements that are both technical and adaptive, thereby requiring combinations of methods to appropriately assess their impact. Machine learning techniques can now be deployed to read textual data at massive scale, thereby enabling – in principle – operational insights from documents such as mid-term and completion reports to be harvested from past and present studies from all over the world, in multiple languages. For small organizations deploying CDD approaches with modest evaluation budgets yet still facing an imperative to assess (and learn from) their efforts, methods such as the Qualitative Impact Protocol (QuIP) (Copestake et al. 2019) could offer a practical alternative. (See also Copestake et al [2021] on using process tracing methods in small-scale impact evaluation, and Guerrero et al [2023] on scaling interventions on the basis of adaptive evaluations.) Meta-analyses and systematic reviews can be useful to the extent they are complemented by approaches identifying more precisely how these impacts and measurements vary with contextual factors. 45 See, for example, the commentary generated by Duncan Green’s (2018) blog post summarizing the findings from a systematic review of CDD projects conducted by 3ie (White et al 2018). 31 Table 7: Example of key research questions and recommended research approach Research question Suggested method(s) or tool(s) 1. How does bundling facilitation with cash • RCT with multiple treatment arms that introduce transfers (at the community or household randomized variation in i) the presence or mode (e.g., level) affect economic welfare, public to household vs. livelihood group) of cash transfers infrastructure, and social sustainability and ii) the presence or type of facilitation. outcomes? 2. Do CDD institutions persist after project • Longitudinal analysis that combines quantitative and completion, and do they generate spillover qualitative data collection from CDD communities impacts (e.g., on service delivery or resilience across multiple projects/countries 46 during the project outcomes)? What elements of design, period and 5+ years after project completion. implementation, and context contribute to • Utilize phone surveys with community these institutional impacts? members/leaders for high frequency and longitudinal feedback. • Where feasible, consider natural or quasi-experiments or equivalence trial method (Burde et al 2023). 3. In what type of contexts can CDD approaches • Qualitative (KIIs, FGDs, ethnography) data collection enhance social cohesion? Which types of social across multiple projects/countries before and during cohesion (linking, bridging, and bonding)? CDD implementation. 47 What core design adjustments are necessary in CDD approaches to create or improve impacts on social cohesion? 4. What approaches to hiring, training, and • Systematic review of current and past CDD project supervising facilitators are adequate/effective implementation documents, including i) to achieve project outcomes, and ii) to Implementation Status and Results Reports (ISR), Mid create spillovers on other activities that Term Reviews (MTR), and Implementation and promote local development? How do Completion Results Reports (ICR). contextual factors affect successful • Use of phone surveys or other digital tools to monitor implementation of this facilitation? How can and assess facilitation at high frequencies over time. small changes to design and implementation • Anthropological process studies. contribute to consistent facilitation? • HR records for basic demographics • RCTs Importantly, the learning agenda requires not only impact evaluations but in-depth studies to understand processes, contexts, and the mechanisms that link project inputs to changes in outcomes. Ideally, such work on understanding the contexts in which CDD projects will work will be conducted before CDD projects are implemented (i.e., as part of the project preparation phase). Knowledge of how local institutions (formal and informal) function; where power lies within communities – the extent to which it is concentrated, and what sources of countervailing power preside; the extent of social cohesion and nature of cleavages; the groups that are marginalized within communities and in what ways. Forging a grounded understanding of each is foundational for the development of a project design that fits with the 46 This approach of looking at similar research questions across several projects/countries would help improve the external validity of results and could draw on lessons from the Metaketa Initiatives. 47 The joint World Bank – Mercy Corps toolkit on social capital and social cohesion provides a qualitative research guide on how to measure social cohesion in CDD projects. See Kim, Sheely, and Schmidt (2020). 32 nature of the contexts in which that project will operate. 48 Table 7 summarizes the need for a variety of research approaches to answer priority research questions on CDD. Studies of implementation processes, and how projects are interacting with local contexts, are also useful during the project’s life. To an extent, World Bank supervision missions aim to play this role, identifying issues and problems that can be addressed. Yet understanding these issues requires more in-depth, usually qualitatively focused, studies that set out hypotheses and systematically test them across a range of contexts. While studies may include non-project ‘control’ sites, the focus is less on identifying impact per se but instead more on teasing out whether or not the assumptions that underpin projects’ theories of change hold up, and if so (or if not) why. Such learning can be useful for operational course adjustments but can also contribute to broader theories on the different ways in which projects affected communities, and why. 5. Conclusion Given the size and scope of CDD investments, and its popularity among governments as a development tool, it is imperative to interrogate CDD’s impacts and ensure best practices are integrated into current and future projects. The World Bank alone supports over 350 CDD or CLD projects in 98 countries with approximately USD 45.2 billion. 49 While there are many other popular development tools, governments often turn to CDD to address overlapping development challenges in difficult environments. As governments have increasingly adopted the CDD approach, often with World Bank financing, the approach has evolved to meet new challenges and opportunities, resulting in projects that include components on devolved financing, household cash transfers, local economic development, climate mitigation and adaptation, and inclusion. This paper calls for a revitalized learning agenda on CDD that focuses on the conditions under which CDD can work and the design and implementation choices that will make CDD most effective within a given context. CDD approaches have a strong record of improving basic services and in some cases also improving economic welfare, participation, and conflict-mediation outcomes. Even as they explicitly harness existing (or establish new forms of) social institutions, however, it is less clear that CDD projects discernably expand social networks and decisively shift pro-social norms – and indeed whether it is desirable that they should singularly expect to be able to do so in such relatively short timeframes. More importantly, many key questions remain unanswered, especially on how variation in project design, implementation modalities, and interactions with contextual factors affect CDD outcomes. Although practitioners often hold tacit knowledge on these factors and incorporate that knowledge into the design and management of their projects, it is important to validate, expand on, and document this knowledge. We identify gaps in the evidence on CDD and outline a research agenda to build knowledge on CDD’s impacts across four outcome categories, based on three sources of variation. We focus on four outcomes tied to social sustainability – cohesion, inclusion, resilience, and process legitimacy – because social 48 Such in-depth analysis is rare. One good example was the Local Level Institutions study in Indonesia that helped shape the design of the Kecamatan Development Project and enabled long-term assessments to be made of social change in everyday village life. See Dharmawan, Jellema, and Wetterberg (2013). A key finding from this study was that conclusions reached after the first round of research were not always supported by research in the second round, raising the obvious specter of whether strident claims and counterclaims based on one-time assessments of interventions in any development sector might be upended if a subsequent round of research were to be conducted. 49 As of July 2023, based on CLD portfolio review. 33 sustainability is critical to development and because, as outlined in Section 4, less is known about CDD’s impacts on social sustainability compared to economic sustainability. However, these are not the only outcomes that matter for CDD nor are they mutually exclusive. We identify three sources of variation that likely contribute to these and other CDD outcomes. This includes variation in project design, implementation approaches, and context. Finally, we call for the use of a diverse range of tools and methods to systematically assess how these sources of variation affect CDD outcomes. This must include standard approaches to project impact evaluations (e.g., RCTs, detailed qualitative and contribution analyses) along with other approaches such as in-depth process evaluations, long-term studies, and the use of administrative and project data. Building this evidence base will improve the design, scalability, and impact of CDD projects and will also generate new knowledge that applies to other similar development approaches. Research that focuses on the ways in which context, design choices, and implementation modalities drive different CDD impacts will help us better understand which elements of the CDD approach are fundamental to achieve impacts and which elements can be adjusted (e.g., the size of a block grant, the type or duration of facilitation) and when. Understanding what designs work and in what contexts will also inform when it makes sense to utilize a CLD approach – for example, an approach where funds are transferred through local governments instead of at the community level (Figure 1). The proposed learning agenda will provide insight on when and how to augment a CDD approach with complementary development activities such as household cash transfers, climate mitigation and adaptation programs, or initiatives focused on women’s economic empowerment. Finally, because certain aspects of CDD projects – such as community consultations – are common in other development approaches, the challenges of assessing the effectiveness of CDD projects outlined in this paper, along with the suggested responses to them, are likely to apply to other areas of development (e.g., field-level extension work in agriculture, facilitation in group-based credit programs). We conclude by acknowledging there are many challenges to implementing this learning agenda. We are not the first to call for more nuance when evaluating CDD’s impacts (see Wong and Guggenheim 2018), which raises the question of why there has not been more progress. We highlight three challenges. First, there is a human resource challenge among operations teams. Most CDD operations teams, at least within the World Bank, are aware of the gaps in the evidence and have a desire to help fill those gaps. However, their priority is on ensuring effective implementation and they rarely have the human resources – team members with technical skills and time – to design or manage research studies, especially studies focused on outcomes not directly tied to the project’s results framework. Second, there is a coordination challenge. Several of the research questions outlined in Section 4 require research designs across multiple projects and countries or over long periods of time. Coordination is also required to identify research opportunities across a large portfolio of operations, to connect researchers to operations teams and data sources, and to integrate the findings generated by researchers from different disciplinary backgrounds. Third, there is a knowledge management challenge of ensuring that new evidence is publicly accessible, disseminated, interpreted for non-technical audiences, and integrated into new project designs. New investments in human resources and a coordinating body would help to alleviate these challenges and support a ‘big push’ forward on the CDD learning agenda. 34 References Alsop, Ruth, Mette Bertelsen, and Jeremy Holland. 2006. “Empowerment in Practice: From Analysis to Implementation.” Washington, DC: World Bank. Anderson, Michael, and Jeremy Magruder. 2017. “Split-Sample Strategies for Avoiding False Discoveries.” No. W23544 National Bureau of Economic Research. Andrews, Matt, Lant Pritchett, and Michael Woolcock. 2017. Building State Capability: Evidence, Analysis, Action. Oxford: Oxford University Press. Atos Consulting. 2014. NSP: Phase III Financial and Economic Analysis Final Report. Washington, DC: World Bank. Arcand, Jean-Louis, Adama Bah, and Julien Labonne. 2011. “Conflict, Ideology and Foreign Aid.” CERDI working paper no. 201021. Arcand, Jean-Louis, and Leandre Bassole. 2008. “Does Community Driven Development Work? Evidence from Senegal.” SSRN Electronic Journal. Arnold, Margaret, Robin Mearns, Kaori Oshima, and Vivek Prasad. 2014. “Climate and Disaster Resilience. The Role for Community-Driven Development.” Social Development Working Paper no 850007. Washington, DC: World Bank. Avdeenko, Alexandra, and Michael J. Gilligan. 2015. “International Interventions to Build Social Capital: Evidence from a Field Experiment in Sudan.” American Political Science Review 109(3): 427-449. Ban, Radu, Michael J. Gilligan, and Matthias Rieger. 2020. “Self-Help Groups, Savings, and Social Capital: Evidence from a Field Experiment in Cambodia.” Journal of Economic Behavior and Organization 180: 174-200. Barron, Patrick. 2011. “Community-Driven Development in Post-Conflict and Conflict-Affected Areas: Experiences from East Asia.” World Development Report background paper. Washington, DC: World Bank. Barron, Patrick and Adam Burke. 2008. Supporting Peace in Aceh: Development Agencies and International Involvement. Policy Studies No. 47. Washington, DC: World Bank. Barron, Patrick, Louise Cord, Jose Cuesta, Sabina A. Espinoza, Greg Larson, and Michael Woolcock. 2023. Social Sustainability in Development: Meeting the Challenges of the 21st Century. Washington, DC: World Bank. Barron, Patrick, Rachael Diprose, and Michael Woolcock. 2011. Contesting Development. Participatory Projects and Local Conflict Dynamics in Indonesia. New Haven: Yale University Press. Barron, Patrick, Macartan Humphreys, Laura Paler, and Jeremy Weinstein. 2009. “Community‐Based Reintegration in Aceh: Assessing the Impacts of BRA‐KDP.” Indonesian Social Development Paper No.12. Washington DC: World Bank. Baseler, Travis, Thomas Ginn, Robert Hakiza, Helida Ogude-Chambert, Olivia Woldemikael. 2023. “Can Redistribution Change Policy Views? Aid and Attitudes toward Refugees in Uganda.” Center for Global Development Working Paper no 645. Washington DC. 35 Beath, Andrew, Fotini Christia, and Ruben Enikolopov. 2013. Randomized Impact Evaluation of Afghanistan’s National Solidarity Program: Endline Report. Washington, DC: World Bank. Beath, Andrew, Fotini Christia, and Ruben Enikolopov. 2017. “Direct Democracy and Resource Allocation: Experimental Evidence from Afghanistan.” Journal of Development Economic 124: 199-213. Beath, Andrew, Fotini Christia, and Ruben Enikolopov. 2018. “Can Development Programs Counter Insurgencies? Evidence from a Field Experiment in Afghanistan.” MIT Political Science Working Paper. Beatty, Amanda, Ariel BenYishay, Elisabeth King, Aniceto Orbeta, and Menno Pradhan. 2018. Kalahi- CIDSS Impact Evaluation Third Round Report. Innovations for Poverty Action (IPA). BenYishay, Ariel, Bradley Parks, Rachel Trichler, Christian Baehr, Daniel Aboagye, and Punwath Prum. 2019. Building on a Foundation Stone: The Long-Term Impacts of a Local Infrastructure and Governance Program in Cambodia. Stockholm: Swedish EBA, AidData, and Open Development Cambodia. Blattman, Christopher, and Eric Green, Julian Jamison, M. Christian Lehmann, and Jeannie Annan. 2016. “The Returns to Microenterprise Support Among the Ultra-Poor: A Field Experiment in Post-War Uganda.” American Economic Journal: Applied Economics 8(2): 35–64. Bold, Tessa, Mwangi Kimenyi, Germano Mwabu, and Justin Sandefur. 2018. “Experimental evidence on scaling up education reforms in Kenya.” Journal of Public Economics 168: 1-20. Boret, Ny, Kishore Gawande, and Daniel P. Kobb. 2021. “Can decentralization lower poverty? Cambodia’s Commune and Sangkat Fund.” World Development 146: 105548. Burde, Dana, Joel Middleton, Roxanne Rahnama, Cyrus Samii. 2023. “Can Communities Take Charge? A Randomized Controlled Trial on Sustaining Schools in Afghanistan.” Working Paper. Callen, Michael, and Miguel Fajardo-Steinhauser, Michael Findley, Tarek Ghani, and Shahim Kabuli. 2023. “Digital Humanitarian Payments to Vulnerable Afghans.” JPAL Policy Report. Casey, Katherine. 2018. “Radical decentralization: does community-driven development work?” Annual Review of Economics 10: 139-163. Casey, Katherine, Rachel Glennerster, and Edward Miguel. 2012. “Reshaping Institutions: Evidence on Aid Impacts Using a Preanalysis Plan.” The Quarterly Journal of Economics 127 (4): 1755-1812. Casey, Katherine, Rachel Glennerster, and Edward Miguel. 2013. The GoBifo Project Evaluation Report: Assessing the Impacts of Community Driven Development in Sierra Leone: Final Report. Washington, DC: International Initiative for Impact Evaluation. Casey, Katherine, Rachel Glennerster, Edward Miguel, and Maarten J. Voors. 2023. “Long Run Effects of Aid: Forecasts and Evidence from Sierra Leone.” The Economic Journal 133(652): 1348-1370. Chase, Robert S., Rikke Nording Christensen, and Maniemai Thongyou. 2010. “Picking Winners or Making Them? Evaluating the Social Capital Impact of CDD in Thailand.” Case Studies in Business, Industry and Government Statistics. 3(2): 1-29. 36 Chatterjee, Shreya, Marine Gassier, and Nikolas Myint. 2023. “Leveraging Social Cohesion for Development Outcomes.” Policy Research Working Paper 10417. Washington, DC.: World Bank. Coleman, James S. 1988. “Social Capital in the Creation of Human Capital.” American Journal of Sociology 94: 95-120. Copestake, James, Marlies Morsink, and Fiona Remnant. 2019. Attributing Development Impact: The Qualitative Impact Protocol Case Book. Practical Action Publishing. Copestake, James, Gary Goertz, and Stephan Haggard. 2021. “The veil of ignorance process tracing (VoiPT) methodology.” Qualitative & Multi-Method Research 18(2): 19-26. Crost, Benjamin, Joseph Felter, and Patrick Johnston. 2014. “Aid under fire: development projects and civil conflict.” American Economic Review 104(6): 1833-1856. de Regt, Jacomina, Shruti Majumdar, and Janmejay Sing. 2013. “Designing Community-Driven Development Operations in Fragile and Conflict-Affected Situations: Lessons from a Stocktaking.” Washington, DC: World Bank. Datta, Upamanyu, Vivian Hoffmann, Vijayendra Rao, and Vaishnavi Surendra. 2015. Report on the Impact of JEEViKA: Evidence from a Randomized Rollout 2011-2014. Washington, DC: World Bank. Datta, Upamanyu. 2016. Socio-Economic Effects of a Self-Help Group Intervention: Evidence from Bihar, India. Washington, DC: World Bank. Deininger, Klaus and Yanyan Liu. 2009. “Longer-Term Economic Impacts of Self-Help Groups in India.” Policy Research Working Paper No. 4886. Development Research Group, Sustainable Rural and Urban Development Team. Washington, DC: World Bank. Dharmawan, Leni, Jon Robbert Jellema, and Anna Maria Wetterberg. 2013. “Studi Kelembagaan Tingkat Lokal Ke-3.” Jakarta: Bappenas. Donahue, D.J. and Gary Glass. 2019. Citizens’ Charter Afghanistan Project Returns and Cost Comparisons: Afghanistan Reconstruction Trust Fund Third Party Monitoring – Supervisory Agent. Washington, DC: Management Systems International. Easterly, William. 2006. “Planners versus Searchers in Foreign Aid” Asian Development Review 23(2): 1- 35. El-Kogali, Safaa, and Caroline Krafft, Touhami Abdelkhalek, Mohamed Benkassmi, Monica Chavez, Lucy Bassett, and Fouzia Ejjanoui. 2016. “The Impact of a Community Development and Poverty Reduction Program on Early Childhood Development in Morocco.” Policy Research Working Paper; No 7671. Washington, DC: World Bank. Ensminger, Jean. 2017. “Corruption in community-driven development: A Kenyan case study with insights from Indonesia.” U4 Issue. Esenaliev, Damir, Aida Bolotbekova, Glzhan Asylbek kyzy, Kanat Tilekeyev, Anastasia Aladysheva, Roman Mogilevskii, and Tilman Brück. 2018. “Social Cohesion through Community-based development in Kyrgyzstan.” Stockholm International Peace Research Institute. Working Paper #46. 37 Fearon, James D, Macartan Humphreys, and Jeremy M. Weinstein. 2009. “Can Development Aid Contribute to Social Cohesion after Civil War? Evidence from a Field Experiment in Post-Conflict Liberia.” American Economic Review Papers and Proceedings 99(2): 287-91. Filmer, Deon, Jed Friedman, Eeshani Kandpal, and Junko Onishi. 2023. “Cash Transfers, Food Prices, and Nutrition Impacts on Ineligible Children.” Review of Economics and Statistics 105(2): 327-343. Fiszbein, Ariel and Norbert Schady. 2009. Conditional Cash Transfers: Reducing Present and Future Poverty. Washington, DC: World Bank Publications. Gibson, Christopher and Michael Woolcock. 2008. “Empowerment, Deliberative Development and Local Level Politics in Indonesia: Participatory Projects as a Source of Countervailing Power.” Studies in Comparative International Development 43(2): 151-180. Guerrero, Isabel, with Siddhant Gokhale, and Jossie Fahsbender. 2023. Scaling Up Development Impact Bowker. Guggenheim, Scott. 2021. “Putting People First in Practice: Indonesia and the Kecamatan Development Program,” in Maritta Koch-Wieser and Scott Guggenheim (eds.). Social Development in the World Bank: Essays in Honor of Michael M. Cernea. New York: Springer. Haines, Rebecca. 2020. Social Inclusion in Fragile Contexts: Pathways Toward the Inclusion of Women in Local Governance Processes: Perspectives from Afghanistan. Geneva: CARE International. Hakiman, Kamran and Ryan Sheely. Forthcoming. “Unlocking the Potential of Participatory Planning: How Flexible and Adaptive Governance Interventions Can Work in Practice” Studies in Comparative Economic Development Heß, Simon, Dany Jaimovich, and Matthias Schündeln. 2021. “Development Projects and Economic Networks: Lessons from Rural Gambia.” Review of Economic Studies 88(3): 1347-1384. Hirschman, Albert O. 1967. Development Projects Observed. Washington, DC: The Brookings Institution. Humphreys, Macartan, Raul Sanchez de la Sierra, and Peter van der Windt. 2012. Social and Economic Impacts of Tuungane: Final Report on the Effects of a Community Driven Reconstruction Program in Eastern Democratic Republic of Congo. New York: Columbia University. IFAD (International Fund for Agricultural Development). 2020. Community-driven development in IFAD- supported projects: Evaluation synthesis. Washington, DC: IFAD. Javed, Amna, Najaf Zahra, Ana Maria Munoz Boudet. 2022. “What do we know about interventions to increase women’s economic participation and empowerment in South Asia?” South Asia Gender Innovation Lab. Kalyvas, Stathis 2006. The Logic of Violence in Civil War. Cambridge University Press. Khanna, Madhulika, Nishtha Kochhar, and Nethra Palaniswamy. 2015. "A retrospective impact evaluation of the Tamil Nadu empowerment and poverty alleviation (Pudhu Vaazhvu) project." The Journal of Development Studies 51(9): 1210-1223. Kim, Jeeyon, Ryan Sheely and Carly Schmidt. 2020. Social Capital and Social Cohesion Measurement Toolkit for Community-Driven Development Operations. Washington, DC: Mercy Corps and The World Bank. 38 Kochar, Anjini, Bidisha Barooah, Chandan Jain, Geeta Singh, Nagabhushana Closepet, Raghunathan Narayanan, Ritwik Sarka, and Rohan Shah. 2020. Impact Evaluation of the National Rural Livelihoods Project. 3ie Impact Evaluation Report 128. New Delhi: International Initiative for Impact Evaluation (3ie). Koch-Wiesser, Maritta and Scott Guggenheim (eds.). 2021. Social Development in the World Bank: Essays in Honor of Michael M. Cernea. Springer. Krishna, Anirudh, Daimon Kambewa, Frank Tchuwa, Frank Kasonga, and Patrick Higdon. 2024. “What do Communities Feel About Community-Driven Development? Learning From Investigations in Rural Malawi.” Duke University, mimeo. Labonne, Julien. 2013. “The KALAHI-CIDSS Impact Evaluation: A Revised Synthesis Report.” Sustainable Development Discussion Paper 69094. Washington, DC: World Bank. Labonne, Julien and Robert S. Chase. 2008. “Do Community-Driven Development Projects Enhance Social Capital? Evidence from the Philippines.” Policy Research Working Paper No. 4678. Washington, DC: World Bank. Macours, Karen and Renos Vakis. 2014. “Changing Households’ Investment Behaviour through Social Interactions with Local Leaders: Evidence from a Randomized Transfer Programme.” The Economic Journal. 124 (576): 607-633. Majumdar, Shruti and Vijayendra Rao. 2017. “On the Frontlines of Scaling-Up: A Qualitative Analysis of Implementation Challenges in a CDD Project in Rural India.” Washington, DC: World Bank Policy Research Working Paper No. 8039. Mansuri, Ghazala and Vijayendra Rao. 2012. Localizing Development. Does Participation Work? Washington, DC: World Bank. Millán, Teresa Molina, Tania Barham, Karen Macours, John A. Maluccio, and Marco Stampini. 2019. “Long-term impacts of conditional cash transfers: Review of the evidence.” World Bank Research Observer 34(1): 119-159. Mvukiyehe, Eric and Peter van der Windt. 2020. “Assessing the Longer-Term Impact of Community- Driven Development Programs: Evidence from a Field Experiment in the Democratic Republic of Congo.” Policy Research Working Paper No. 9140. World Bank, Washington, DC. Neelim, Ananta and Joseph Vecci. 2013. Evaluation of the Solomon Islands Rural Development Program. Clayton: Monash University. Nkonya, Ephraim, Dayo Phillip, Tewodaj Mogues, John Pender, and Edward Kato. 2012. “Impacts of community-driven development programs on income and asset acquisition in Africa: The case of Nigeria.” World Development 40 (9): 1824-1838. Olken, Benjamin A. 2007. “Monitoring Corruption: Evidence from a Field Experiment in Indonesia.” Journal of Political Economy 115 (2): 200-49. Olken, Benjamin A. 2010. “Direct Democracy and Local Public Goods: Evidence from a Field Experiment in Indonesia.” American Political Science Review 104 (2): 243-267. 39 Olken, Benjamin A., Junko Onishi, and Susan Wong. 2011. Indonesia's PNPM Generasi Program: Final Impact Evaluation Report. Jakarta: World Bank. Paler, Laura, Cyrus Samii, Matthew Lisiecki, and Adrian Morel. 2015. Social and Environmental Impact of the Community Rangers Program in Aceh. World Bank. Washington, DC: World Bank. Pandey, Vivek, Abhishek Gupta, and Shivani Gupta. 2019. “Labor and Welfare Impacts of a Large-Scale Livelihoods Program.” World Bank Policy Research Working Paper No. 8883. Washington, DC: World Bank. Parajuli, Dilip, Gayatri Acharya, Nazmul Chaudhury, and Thapa Bishnu Bahadur. 2012. Impact of Social Fund on the Welfare of Rural Households: Evidence from Nepal Poverty Action Fund. Washington, DC: World Bank. Parks, Bradley C., Matthew DiLOrenzo, and Daniel Aboagye. 2019. “Closing the Broken Feedback Loop: Can Responsive Aid Strengthen State Legitimacy?” AidData Working Paper 73. Williamsburg, VA: AidData at William & Mary. Parthasarathy, Ramya, Vijayendra Rao, and Nethra Palaniswamy. 2019. “Unheard Voices: The Challenge of Inducing Women’s Civic Speech.” World Development 115: 64-77. Pomeroy, Jacqueline. 2016. The Economic Impacts of PNDS Infrastructure Projects. A study prepared for the Ministry of State Administration and the Ministry of Planning and Strategic Investment. Dili: Government of Timor Leste. PRIMEX. 2016. Technical Audit and Economic Analysis for the KALAHI-CIDSS National Community-Driven Development Program: Final Report. Manilla: PRIMEX. Pritchett, Lant and Michael Woolcock. 2004. “Solutions When the Solution is the Problem: Arraying the Disarray in Development.” World Development 32(2): 191-212. Putnam, Robert. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton, NJ: Princeton University Press. Rao, Vijayendra, Kripa Ananthpur, and Kabir Malik. 2017. “The Anatomy of Failure: An Ethnography of a Randomized Trial to Deepen Democracy in Rural India.” World Development 99(11): 481-497. Rawlings, Laura, Lynne Sherburne‐Benz, and Julie van Domelen. 2003. Evaluating Social Funds: A Cross Country Analysis of Community Investments. Washington, DC: World Bank. Samii, Cyrus. 2023. “Revisiting community-driven reconstruction in fragile states.” WIDER Working Paper 2023/26. Sanyal, Paromita, Vijayendra Rao, and Shruti Majumdar. 2015. “Recasting Culture to Undo Gender: A Sociological Analysis of Jeevika in Rural Bihar, India.” Policy Research Working Paper No. 7411. Washington, DC: World Bank. Singh, Pramod, Harekrishna Misra, Vivek Pandey, H. Shylendra, Rakesh Arrawatia, Shambu Prasad, Ram Vikas, Mukul Kumar, and Sankar Datta. 2017. Independent Assessment of Design, Strategies, and Impacts of DAY-NRLM. Anand, India: Institute of Rural Management Anand. Singh, Pramond and Vivek Pandey. 2019. Independent Assessment of DAY-NRLM in Jammu and Kashmir. Anand, India: Institute of Rural Management Anand. 40 The Asia Foundation. 2018. Citizens’ Charter “Booster” to the 2018 Survey of the Afghan People: Baseline Report. Kabul: The Asia Foundation. Voss, John. 2012. PNPM Rural Impact Evaluation. Washington, DC: World Bank. Wetterberg, Anna, Jon R. Jellema, and Leni Dharmawan. 2012. The Local Level Institutions Study 3: Overview Report. Jakarta: PNPM Support Facility. White, Howard, Radhika Menon, and Hugh Waddington. 2018. "Community-driven development: Does it build social cohesion or infrastructure? A mixed-method evidence synthesis." International Initiative for Impact Evaluation, Working Paper No. 30. Winkler, Stephen and Camila Linneman. 2022. “Evidence note on community driven development (CDD): A summary of key findings on World Bank financed CDD operations.” Washington, DC: World Bank. Wong, Susan. 2012. “What Have Been the Impacts of World Bank Community Driven Development Programs? CDD Impact Evaluation Review and Operational and Research Implications.” Social Development Department Report. Washington, DC: World Bank. Wong, Susan, and Scott Guggenheim. 2018. "Community-driven development: myths and realities." Social, Rural, Urban Resilience Global Practice, World Bank Policy Research Working Paper 8435. Washington, DC: World Bank. Woolcock, Michael. 2019. “When do Development Projects Enhance Community Well-Being?” International Journal of Community Well-Being 2(2): 81-89 Woolcock, Michael. 2022. "Will it work here? Using case studies to generate ‘key facts’ about complex development programs." In Jennifer Widner, Michael Woolcock, and Daniel Ortega Nieto (eds.) The Case for Case Studies: Methods and Applications in International Development, New York: Cambridge University Press, pp. 87-116. World Bank. 2004. World Development Report 2004: Making Services Work for Poor People. Washington, DC: World Bank. World Bank. 2008. Community-Driven Approaches in Lao PDR: Moving Beyond Service Delivery – Summary Overview. Washington, DC: World Bank. World Bank. 2011. World Development Report 2011: Conflict, Security, and Development. Washington, DC: World Bank. World Bank. 2012. Afghanistan – Second Emergency National Solidarity Project: Implementation Completion Report and Results Report. Washington, DC: World Bank. World Bank. 2013. Indonesia: Evaluation of the Urban Community Driven Development Program, Program Nasional Pemberdayaan Masyarakat Mandiri Perkotaan. Washington, DC: World Bank. World Bank. 2016a. A study on the economic impacts of PNPM Rural 2012-15. Washington, DC: World Bank. World Bank. 2016b. Lao Poverty Reduction Fund Phase II Impact Evaluation: Final Report. Washington, DC: World Bank. 41 World Bank. 2021a. Disclosable Version of the ISR – Citizens’ Charter Afghanistan Project – P160567 – Sequence No: 09. Washington, DC: World Bank. World Bank. 2021b. Disclosable Version of the ISR – Poverty Reduction Fund III – P157963 – Sequence No. 09. Washington, DC: World Bank. World Bank. 2023. Moving from Traditional to Transformational Approaches to Women’s Economic Empowerment. World Bank. Washington, DC: World Bank. Zurstrassen, Matthew. 2020. A Review of the Myanmar National Community Driven Development Project in Conflict-Affected Contexts. Washington, DC: World Bank. 42