Leveraging Social Cohesion for Development Outcomes

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.


INTRODUCTION
Social cohesion in its broadest sense refers to the ability of communities and individuals to collaborate with one another and with government for the common good.Strong social cohesion makes communities more resilient to external shocks and better able to manage public goods, which in turn facilitate stability and growth.Conversely, gaps in social cohesion can substantially exacerbate development challenges by undermining the ability of communities to overcome conflicts in a non-violent, productive manner and to act collectively to identify and implement solutions to the issues they face.
Social cohesion takes on particular importance in situations of fragility, conflict, and violence (FCV), where limited state capacity often meets extensive urgent needs.In such situations, the ability of governments and development partners to work directly with communities in identification of needs and vulnerable populations as well as in the implementation and supervision of development programs offers an important potential asset.Ten years ago, the 2011 World Development Report on Conflict, Security and Development included programs that support bottom-up state-society relations as the first in its top five policy recommendations for programming in insecure areas, and specifically highlighted the importance of social cohesion as part of national institutional capacity.Similarly, the 2018 UN-World Bank Pathways for Peace study argued that "navigating change by fostering inclusiveness and thus social cohesion is the essence of prevention."This theme was reiterated in the WBG's FY20-25 FCV Strategy, which highlights the importance of fostering social cohesion as a core engagement principle in FCV contexts.Beyond FCV settings, social cohesion enables communities to respond to other shocks, such as responding to the COVID-19 pandemic or adapting to the impacts of climate change.
Furthermore, in a recent agenda-setting strategy paper,2 the World Bank identified social cohesion as one of the key pillars of social sustainability, along with inclusion, resilience, and process legitimacy.The dynamics connecting these different pillars are complex, and efforts to promote one component of sustainability can create disruptions that, at least initially, undermine its other dimensions, as is the case when interventions in favor of greater inclusion erode cohesion.Nevertheless, with careful adaptation to local challenges, efforts to advance these four components can also be mutually reinforcing.The strategies outlined in this paper aim to strengthen social cohesion while also creating contexts that allow communities to become more inclusive and resilient, and in which negotiation the process between different stakeholders is seen to be legitimate.Achieving these mutually reinforcing gains can however be a slow process, and development actors enacting these approaches must recognize that short-term engagement is often insufficient.
Efforts to operationalize social cohesion in development projects have had to contend with multiple definitions of the term, a lack of clarity on diagnostic and measurement approaches, and contradictory evidence on the feasibility of influencing social cohesion through project level interventions.This framing paper aims to address some of these constraints by offering a definition of social cohesion grounded in the academic literature and aligned with the institutional and operational priorities of the World Bank, and by bringing together new evidence and literature in the field to offer pragmatic approaches for measuring and seeking to build social cohesion in the context of development interventions.
The objective of the paper is to identify how to foster and leverage social cohesion to improve development outcomes through a review of global evidence and literature.While grounded in theory and evidence from academia, the paper is aimed at an audience of development practitioners, including in the World Bank as well as within national governments, development partners, and civil society.
In exploring social cohesion, the paper focuses on communities as the primary unit of analysis.Trends in social cohesion can be examined at different scales, with a growing literature on the determinants of social cohesion at the national level.However, such approaches tend to mask significant subnational differences and therefore provide insufficient granularity to offer operational entry points.This is especially true in FCV contexts, where the conflict literature highlights the importance of the interaction of national and local dynamics in escalation and violence (Kalyvas 2006).To complement existing frameworks largely focused on national-level drivers of fragility, this paper therefore examines community-level dynamics and national-local interactions.In doing so, it also highlights the potential of local participatory approaches to strengthen cohesion.
The structure of the paper is as follows: section 1 defines social cohesion, provides a historical and development context of the term and its use, and identifies its key constituent components and relationships.Section 2 explores challenges in measuring social cohesion, reviewing potential diagnostic tools and measurement approaches.Section 3 reviews the evidence base on approaches to building social cohesion.Section 4 concludes with key takeaways as well as areas for future research.

DEFINING SOCIAL COHESION
Social cohesion is a term with a long history and a subject of inquiry across multiple disciplines, leading to a range of definitions.While these vary in focus and have evolved over time, an emerging consensus in today's literature conceptualizes social cohesion as the ability of people to cooperate for a common good.This section reviews the history and evolution of the term, proposes a specific definition for the purposes of this paper and related work by the World Bank, and reviews the relationship of the term to concepts such as social capital.Clarity around the concept of social cohesion provides the foundations for effective diagnostics, operational interventions and measurement of outcomes that are the focus of the remainder of this paper.

SOCIAL COHESION IN HISTORIC CONTEXT
When first conceptualized, social cohesion was most frequently identified by its absence and its resultant consequences, including conflict.In 1897, French sociologist Emile Durkheim first defined social cohesion in the academic literature, describing it as a societal characteristic of interdependence between individuals and defining it as the absence of latent social conflict and the presence of strong bonds.In 1909, American sociologist Charles Cooley coined the idea of primary groups, identified as groups having communication, cooperation, and a high number of friendships stemming from time together, which when absent can cause social disorganization.
There was subsequently more headway in identifying what social cohesion independently looks like and what characterizes it.Further studies identified varying dimensions, such as specific characteristics that affect group dynamics (French, 1959); an intrinsic collective mentality with reciprocity, common ways of thinking and feeling (McDougall, 1921); and as rewards from a set of negotiated exchanges in people's friendships, with more exchanges leading to more social cohesion (Homan, 1958).
Over the past 30 years, the focus turned to social cohesion as a prescriptive process, emphasizing the need for inclusion and participative structures that bring together institutions and citizens.Defined as an ongoing process, socially cohesive societies are those that reduce disparities in wealth and income, and have mechanisms to solve and navigate conflict (Beauvais, 2002), fight exclusion and marginalization, and offer opportunities and participatory processes to improve social mobility (OECD, 2008).Aside from these actionable elements, the definitions over time have stressed and reached a consensus on the importance of shared values, identities, and communities of interpretation (Maxwell 1996), as well as on levels of solidarity, reciprocity, and trust, including through strong social norms, pressures to conform and mechanisms through which norms are enforced.
Schiefer and Van der Noll's (2017) comprehensive literature review identified three critical dimensions of social cohesion.Firstly, socially cohesive societies are characterized by close social relationships through strong social networks, high levels of institutional and social trust and civic participation in public life.Secondly, members of society are emotionally connected to the social identity, and thirdly, exhibit an orientation towards the common good and solidarity and compliance with social order.Leininger et. al (2021) further adopt a lean definition of social cohesion emphasizing vertical and horizontal relations among members of society and state which hold together and characterize social cohesion as a set of attitudes and behavioral manifestations.These behaviors include trust, inclusive identity, and cooperation for common good.Inclusive identity refers to social identities or hierarchical memberships in different groups, where inclusion is greatest when superordinate social identities create inclusive compatibility between various subordinate identities without dominating them.Trust refers to social trust and not particularized trust.Cooperation for common good refers to the manifestation of cooperation and commitment and not the intent or attitude.These attitudes then build upon and reinforce each other, wherein a minimum level of trust and inclusive identity is a precondition for cooperation.

AN OPERATIONAL DEFINITION OF SOCIAL COHESION
Across this variety of perspectives, a consensus is emerging on what the hallmarks of strong, productive social cohesion are.These include high levels of trust, openness towards "outgroups", and the perception that public institutions are legitimate -all features of communities that have greater capacity for effective collective action.Based on this emerging consensus, we adopt the following definition: Social cohesion is a sense of shared purpose, trust and willingness to cooperate among members of a given group, between members of different groups, and between people and the state.
What constitutes an individual's community is context-dependent and has evolved over time.With the evolution of national institutions and changes in communication technologies, scholars coined the phrase "imagined communities" to refer to communities (including nations) which exist only "in the minds" of their members, most of whom have never met or even heard of each other (Anderson, 1983).But while these imagined communities may play an increasingly important role in shaping individuals' social and political behaviors, they have not become the only form of community that informs such behavior.The propensity of individuals to act collectively, as well as the types and direction this collective action may take also depends on the more immediate communities they are a part of.These communities may be constituted through proximity, history, and regular interaction in a social or professional context.They may result from shared material interests, including the need to manage shared resources (for example villages sharing a watershed).They may also be composed of individuals who, even if they do not know each other, are connected by kinship or other ties (for example extended clan structures).
These more immediate communities vary greatly in their nature, history, as well as in how explicit, inclusive and participatory the norms governing them are.Migdal's conceptualization of state-society relations, which shows how the state competes with a range of other entities for social control through "ongoing struggles among shifting coalitions about the rules for daily behaviors," offers a helpful template for identifying the overlapping structures of collective action and communities that may operate in a given environment.To identify such communities, it is important to look for the structures that have "the ability to devise rules governing aspects of people's lives" and may "offer individuals strategies or survival, and, for some, strategies of upward mobility."This analysis further points to three markers that signal a community's influence over its members: compliance with the norms it sets and/or the demands of its leadership, participation in the activities it organizes, and acceptance of the "rules of the game" it defines.
Given the web of communities that may shape whom people are willing to trust and cooperate with, and the complex ways in which they may reinforce or challenge each other, what is the relevant type or "level" of community that discussions of social cohesion should consider?The answer depends on context, but the framework introduced in this paper points to a set of questions that can guide such discussions.What are the structures in place that support practice of solidarity or non-violent dispute resolution?What are the main threats to such structures?Whom might they exclude or disadvantage?What are the contested resources to be managed?What are the existing networks and norms?
The definition we propose is meant to help navigate these questions by highlighting three sets of relations that contribute to social cohesion: those connecting individuals within a given community (bonding), those connecting individuals across distinct communities (bridging), and those connecting citizens to people or structures in a position of power (linking).The first two sets of relations contribute to horizontal social cohesion, and the third one to vertical social cohesion.
While there are differences in the nature of the horizontal and vertical components, all three dimensions are integral to social cohesion.Unlike horizontal social cohesion, vertical social cohesion is asymmetrical: the state can obtain compliance from individuals in ways that are not available to individuals interacting with one another.A society characterized by strong vertical cohesion is primarily one in which individuals are willing to cooperate with the state, rather than the reverse.In spite of this asymmetry, this vertical dimension remains critical to the definition of social cohesion.This is because a great variety of interactions between members of a group or different groups are mediated and sustained by state institutions.Furthermore, while this vertical dimension is asymmetrical, it is not unidirectional: strengthening social cohesion also means putting people in a better position to hold state institutions accountable.
Vertical social cohesion links all people in a country to state institutions.While the literature commonly uses the phrase "citizen-state relationship" in reference to vertical social cohesion (and we use this phrase as well), strong social cohesion must be inclusive of all individuals in a country, not just nationals of that country.Furthermore, a distinction can be made between trust in state institutions and trust in representatives of a particular government.While the behavior of both may affect the trajectory of a society, the vertical component of social cohesion primarily reflects the former kind of trust.
The three dimensions of social cohesion (bonding/bridging/linking) are connected to other core features of a society, particularly the quality of social capital and of the social contract.All three enable people to act collectively, but in different ways or at different levels.The reciprocal relationships between these three dimensions are further discussed in section 1.4 below.Figure 1 illustrates the operational implications of this conceptualization, highlighting specific issues associated with gaps along each dimension and mechanisms to remedy them.

❑
Ineffective management of shared resources (including land), exclusion of minorities/outgroups from these resources ❑ Gaps in access to services between groups or along identity lines ❑ Tensions/conflictual relations between host communities and displaced people ❑ Inter-community violence This definition is rooted in the academic literature and historical traditions of the term, while offering three distinct strengths.First, it can guide diagnostics and measurement of social cohesion and highlight mechanisms that interventions can aim to strengthen.Second, it is consistent with the characteristics of social cohesion, not only as defined over time by a multitude of scholars, but also as used in current practice by other organizations (for example, it is in line with the definitions of the OECD3 and UNDP4 in highlighting the importance of participatory processes, role of trust and shared purpose and cooperation).Finally, it explicitly introduces the relational element of social cohesion, and identifies key participants in the process of building social cohesion, offering a framework that can guide operational responses.
To understand social cohesion, it is important to consider the multiple levels at which social interactions play out.The proposed definition considers the institutional, community, sub-community, and individual levels, providing a more holistic sense for those looking to measure, foster or leverage social cohesion effectively.
Social cohesion happens at the intersection of three levels -individual, community and institutional, where formal societal structures must allow citizens to act, along with diverse communities with shared norms and values, to enable individuals to affirm their commitment to belong, participate and perform (Xavier, 2019).The relational element of the definition, by specifying horizontal relationships within groups (bonding), across groups (bridging) and between groups and power structures (linking), highlights the importance of interactions in social cohesion and helps a systematic understanding of the multiple underlying relationships that can foster or degrade it.Moreover, by clearly outlining the various groups involved in understanding social cohesion, it is operationally more directional, by identifying a set of key actors and groups to influence in the process of building social cohesion.These groups can look very different (including at a national and subnational level) but there is sufficient flexibility in what they can mean for this to be a starting point for analysis and intervention.

WHY SOCIAL COHESION MATTERS
Social cohesion is an inherently valuable strength for communities and societies to pursue.It also plays an important role in allowing them to meet their development goals, improve social inclusion and become more resilient to violent conflict.Indeed, social cohesion can help communities achieve a variety of development outcomes by enabling the kind of collective action necessary to manage shared resources and make progress towards shared goals.It allows communities to, for instance, consistently manage local infrastructures, establish mutual aid and self-help mechanisms (such as rotating savings groups), adapt to new environmental challenges or adjust to influxes of displaced households.One reason community-driven programs effectively improve basic public services and infrastructures is that they are typically implemented in contexts characterized by high levels of trust and civic engagement by community members (Casey, 2018).Strong social cohesion also creates an environment that facilitates economic investments and exchanges within and between communities.It does so by fostering the trusts required for agreements to be upheld, and letting communities avoid the economic losses associated with discrimination.
Because social cohesion also reduces the barriers to participation in economic, political and social life that different groups of individuals may face -including women, unemployed youth or minorities -it is a critical ingredient of inclusive societies and communities.As such, it contributes to economic growth and poverty reduction.A growing body of evidence shows that polarization -which undermines linking across identitybased or economic groups -is associated with greater difficulties implementing redistributive policies (Montalvo and Reynal-Queyrol 2003;Alesina and La Ferrara, 2005;Stichnoth and Van der Straeten 2011).
In addition, poor social cohesion and inclusion exacerbate the risk of violent escalation of the conflicts that divide communities.The adoption of exclusionary policies has been shown to predict civil war recurrence, possibly because it undermines trust across communities (Call 2012).Similarly, high levels of inequality, when unaddressed, leave societies more vulnerable to violent conflicts.Conversely, participative governance and greater political engagement from previously marginalized groups can help mitigate the risk of conflict, particularly during economic downturns (Fearon 2010, Fearon and Laitin 2013, Min et al. 2017).At the community level, greater social cohesion can increase resilience to the influence of violent actors, from insurgent movements to criminal organizations.In different contexts, including Colombia or the Philippines, studies have shown that communities' ability to resist the demands of armed groups depends on their members' capacity for effective collective action and the quality of existing local institutions, that is on both the bonding and linking dimensions of social cohesion (Arjona, 2016;Rubin, 2018).
Finally, because, as noted, social cohesion is also a desirable end in and of itself, it is important to note the reciprocal relationships between social cohesion and the different variables considered in this discussion.The lack of social cohesion not only undermines policies meant to address inequality, it also becomes more severe as inequality grows (this relationship has been studied mostly across European countries -see Vergolini, 2011, or Whelan andMaitre, 2005).Similarly, poor social cohesion not only weakens communities' resilience to the threats posed by violent actors, it is also a frequent consequence of violent episodes, which can be difficult for communities to overcome even long after the end of these episodes (Alesina and La Ferrara, 2002).
At the same time, social cohesion is not an unambiguous good.Indeed, communities can be both highly cohesive and organized around inequitable hierarchies.Strong social ties, supported by a sense of solidarity and reciprocal obligation between community members can help armed insurgencies recruit and flourish (Staniland 2014).In some contexts, high levels of bonding can undermine the bridging dimension of social cohesion, with members of highly cohesive communities remaining mistrustful of members of out-groups.Highly cohesive communities may also offer less space and support for those that fall outside traditional norms or practices, including based on gender identity and sexual orientation.Analyses of the genealogy of social contracts also show how exclusionary the political and social processes that produced them can be, depriving racial minorities or other vulnerable groups from the rights and protections that these contracts offered to members of the dominant group (Mills 1997, Goldberg 2002).

CONTEXTUALIZING SOCIAL COHESION WITHIN RELATED CONCEPTS
In describing relationships across society, social cohesion is related to other aspects of socio-political dynamics, including social capital, social resilience, and social contracts.In this section, we examine definitions of these concepts and how they relate to social cohesion.

SOCIAL CA P ITA L
Social capital refers to the norms and networks that enable people to act collectively (Woolcock and Narayan, 2000).While often used synonymously with social cohesion, it is broadly recognized as a critical subcomponent of social cohesion.Initial conceptualizations of social capital focused on its positive manifestations in communities -such as local and civic organizations, where a higher density represented higher levels of social capital.The networks view stressed horizontal and vertical community linkages, while with acknowledging some of the possible negative aspects of group dynamics.(for example strong intra community but weak intercommunity ties could encourage narrow sectarian interests).The institutional view considered that political, legal and institutional environments influence community networks.Finally, the synergy view, which integrates the institutional and networks views, acknowledged that states, firms, and communities are not singlehandedly capable of promoting broad-based sustainable development, and that complementarities and partnerships forged within and across them are therefore critical.This view is based on the concept of complementarity or mutually supportive relations, facilitated through supportive overarching formal and informal institutions and embeddedness or the extent of ties connecting citizens and public officials.
A key distinction between social capital and social cohesion is that social capital generally focuses more on the level of individuals, with an emphasis on networks between and across individuals and the strength of these ties, whereas social cohesion focuses more on society-level manifestations (King et al., 2010;Dayton-Johnson, 2003).Social capital includes an individual's sacrifices made in an effort to cooperate with others.In contrast, social cohesion is a societal characteristic that depends on social capital -more socially cohesive societies will include communities with higher manifestations of social capital -both horizontally and vertically (Oxoby, 2009).Nonetheless, the precise boundaries between social capital and social cohesion are blurry, and have at times been used interchangeably (Mansuri, et al., 2004), including by the World Bank (Beauvais, 2002).
The concept of bonding, bridging, and linking used to organize this paper's discussion of the different dimensions of social cohesion, was first introduced in the context of describing different forms of social capital (Woolcock et al., 2000;Putnam, 2000).In its original usage, bonding social capital refers to trusting and cooperative relations between members of a network who see themselves as being similar.Bridging social capital, by contrast, comprises relations of respect and mutuality between people who know that they are not alike in some socio-demographic sense.Woolcock (2004) further refines the category of bridging, by introducing the subset of linking -defined as norms of respect and networks of trusting relationships between people who are interacting across explicit, formal, or institutionalized power or authority gradients in society.
Over time, this conceptualization of social capital theory has expanded to social cohesion.Colletta et al. (2000) use the concepts of bonding and bridging not only in the context of social capital, but also in the context of broader social relations, using case studies from Rwanda and Cambodia to understand the nexus between violent conflict, social capital and social cohesion and identify social capital as a subset of social cohesion, drawing from the work of Kawachi et al. (2000).King et al. (2010) draw a contrast between interpersonal social cohesion and intergroup social cohesion, acknowledging the resemblance of these concepts with bonding and social capital and articulate their preference of using the term social cohesion to emphasize their focus on attributes of groups and patterns of cooperation directly, rather than assets such as religious beliefs or altruistic dispositions that may drive patterns.
For this framing paper too, communities are the relevant primary unit of analysis and extrapolating the social capital terminology in the context of social cohesion is useful in terms of understanding how specific types of social capital relate to social cohesion.This includes, for instance, what the exclusion of certain groups within a community means for social cohesion, as well as how dynamics at the community level differ when considering cohesion within a community, across communities, or relations with state institutions.

SOCIAL RE S I LIE N CE
Social resilience enhances the capacity of individuals, groups and organizations to deal with threats and is recognized as a key outcome of social cohesion.It is the capacity of actors to access networks and connections in order to not only cope with and adjust to adverse conditions (reactive capacity), but also to search for and create options and thus develop increased competence in dealing with a threat (Obrist et al. 2010a).A first dimension of social resilience is the coping capacities of actors and their ability to overcome adversity (adaptive capacity).A second important dimension rests on actors' ability to learn and adapt to new situations and transformative capacities, to be more societally robust to crisis (Keck et al., 2012).
Social resilience is both an outcome and a driver of social cohesion, which itself depends on social relations and network structures, institutions and power relations, as well as and knowledge and discourse.The presence of social networks, structures and relations alone, however, does not guarantee the building and maintenance of social resilience -it is the content of social relations and the role of trust, reciprocity and mutual support that influences societal capacity to manage change (Peiling et al., 2005).Institutions, another component of social cohesion, also influence social resilience to the extent that they shape social and economic systems in terms of structures and distribution of assets (Adger, 2000).As a subset of social cohesion, social capital is recognized as playing a key role in building and maintaining social resilience.

SOCIAL C O N T R A C T
The "social contract" is the set of often implicit agreements between societal groups and the state on rights and obligations toward each other (Loewe et al., 2021).Conceptually, strong social contracts are both a critical contributor to social cohesion, and one of its outcomes.High levels of social cohesion can help establish stronger social contracts.The elements of social cohesion such as reciprocity, shared values and trust, including within groups and between groups and the state are those that shape a social contract.A weak social contract poses a strong risk to social cohesion (Razavi et al., 2020).For instance, the absence of a strong social contract between a government and its citizens can create the space for rebel and armed groups to step in and provide the expectations of the contract themselves, subverting state legitimacy and creating discord with weaker groups.Alternatively, social cohesion can help build and rebuild social contracts, especially in conflict settings (Kaplan, 2017).

MEASURING SOCIAL COHESION
As the conceptual understanding of social cohesion has evolved, so have the approaches deployed to measure it.This section aims to provide guidance and identify resources to design a strategy for measuring social cohesion at a project level.Given the need for approaches to be grounded in context, there is no one size fits all approach -instead, the focus in this section is on identifying what diagnostic and measurement approaches suit the context and purpose of specific interventions, and most relevant tools and sources for those.It is by its nature not comprehensive, as additional data sources may be or become available or new measurement methods developed.Rather, it is intended as a starting point to ground discussions of measuring social cohesion.
Despite a growing number of data sources, measuring social cohesion remains challenging because it involves high levels of abstraction and the integration of multifaceted constructs.Beyond a sound conceptualization of social cohesion, its measurement requires an understanding of its practical manifestations and of the different categories of actors involved in these manifestations.To be adequately measured, social cohesion must also be contextualized.Measurement strategies must be adapted to different countries, cultures, or socio-economic environments.Tracking the evolution of different dimensions of cohesion can be necessary to better understand how communities are faring: for instance, trust in state services remain unchanged, even while attitudes towards outgroups deteriorate.Tracking both indicators over time is necessary to understand changes in overall social cohesion.
When developing a project level monitoring and evaluation strategy, it can be helpful to think of approaches to measuring social cohesion as falling along a continuum covering diagnosis, monitoring anticipatory analytics developments, and evaluation.Indicators and methods must be adapted to each of these dimensions.The adequacy of indicators -e.g.their quality and temporal coverage, along with the frequency of collection and the context within which they are collected -may change depending on their intended use: diagnostics require credible data, over time and across countries, and specific understanding of the local context.Monitoring requires an understanding not only of the intervention at hand, but also which indicators most closely track anticipated changes.Anticipatory analysis requires identifying key elements that precede changing situations in conflict regions and the ability to collect and analyze them quickly.Evaluation requires recognition of consistent indicators that are responsive to interventions and can be collected over time.

MEASURING SOCIAL COHESION: TOOLS AND DATA SOURCES
There is a myriad of existing and emerging quantitative and qualitative tools to measure social cohesion that draw from research methodologies from economics, behavioral sciences, sociology, political science, and anthropology.Different tools can be employed at different project stages, depending on the purpose of measurement.Some tools are helpful in setting and informing practitioners of the context in which their program will function, while others are useful in diagnosing challenges or formulating hypotheses on the issues a program will address, monitoring the outcomes of a program, predicting future conflicts, or evaluating program effectiveness.No specific tool offers a clear pathway to the right answer.Combining several tools is often necessary.It an offer the flexibility required to adapt to a range of practical and political constraints (including time and budget constraints).It also helps validate initial insights and triangulate findings.For instance, while existing surveys may help identify an issue between ethnic groups in a community, focus groups or interviews may aid practitioners in identifying and confirming a more nuanced understanding of inter-group relations and potential entry points.Figure 3 summarizes available tools and their usefulness across the measurement continuum.The next section includes a discussion on key quantitative and qualitative tools, what they can adequately measure, their appropriateness for different measurement purposes, and associated challenges.

X INDICE S R E L A T E D T O SO C I A L CO H E S I O N P E R C E P T I O N S
Composite indices that integrate multiple dimensions of social cohesion are a widely used tool of measurement.These indices use multiple indicators that capture perceptions and attitudes using data collected via primary surveys designed to track respondents' perceptions of and attitudes towards intra-and inter-community dynamics, as well as local institutions.Many will combine questions regarding opinions or values with hypothetical situations.Indicators capturing intra-community, inter-community and citizen-state dynamics are then aggregated to develop a composite understanding of the state of social cohesion in a specific context.The data for these exercises are often already widely available and already collected, and when not available, there are pre-designed tools of data collection available which can be deployed or customized for projects.

EXISTING SOURCES
A number of survey-based indices tracking social cohesion indicators exist at the global, regional and in some cases national level.At the global level, two of the best-known indices are the World Values Survey and Gallup polls, which are value-based representative surveys capturing data on social capital, trust, organizational membership, perceptions of corruption, ethical and religious values, political participation, and political culture.In addition, the regional barometers (Africa, Latin America, Asia etc.) capture perceptions of state institutions, local government, justice system, security forces, corruption, or access to services, as well as attitudes towards out-groups.
The questions asked by these widely available surveys help understand different dimensions of social cohesion, including bonding, bridging, and linking.For instance, the World Values Survey asks respondents whether they feel insecure or experience racism within their own neighborhood, which attempts to capture dynamics within communities by specifying the geographical parameter for the respondent's consideration.The Afro barometer, asks, "generally speaking, would you say that most people can be trusted or that you must be very careful in dealing with people?".While the question does not explicitly ask for attitudes towards out-groups, Delhey, Newton and Welzel (2011) found that in most countries respondents interpret "most people" as outgroups and the indicator as a result is often used to capture this element of social cohesion.
In addition to existing surveys, there are also predesigned tools that facilitate primary data collection at a community level.The toolkit jointly developed by Mercy Corps and the World Bank includes a set of questions designed to capture the underlying constructs of bonding, bridging, and linking in the context of communitydriven development programs.The toolkit includes questions to understand relationships, resources, trust, collective action norms, civic engagement, identity, belonging and even specific attitudes towards out-groups.
The UNDP Social Cohesion and Resilience Index was developed to address the dearth of perception-based data in FCV contexts.Now used in countries including South Sudan, Nepal, the Maldives, Liberia, and Ukraine, it was initially developed for use in Cyprus to track trust in institutions, human security, and satisfaction with civic life.It includes data on the capacity of public institutions, perceptions of institutional corruption, negative stereotypes, anxiety about interactions with adversarial groups, social distance (weak ties) and threats, and active discrimination towards others.

ADVANTAGE S
Using existing indices is often appealing for purposes of analysis and diagnostics, because data is widely available across countries and collected at frequent intervals.The World Values Survey for instance, has been conducted since 1981 at regular intervals for a wide set of countries that has continued to expand -in 2020, it was conducted in 80 countries.Addressing data gaps in developing countries, the regional barometers cover many countries within their regions, capturing regionally meaningful variation and allowing for regional-level contextualization of the questions asked.For instance, the Afro barometer covered 37 countries in Africa in its sixth round in 2021, was conducted in 115 languages and represented 76 percent of the African population.Given that surveys are conducted at regular intervals, they allow for meaningful comparison over time and across countries -both useful dimensions to consider in social cohesion diagnostics.

CONSTRAINTS
While these indices contain a wealth of data, it is important not to rely solely on a composite index to avoid missing key nuances or changes in components of social cohesion.For example, a country or region may show an improvement in social cohesion overall, but this difference may be driven by an improvement in people's perceptions of state services and miss out on growing polarization among dominant and minority groups.
In addition, it is important to remember that these surveys are not specifically geared to capture social cohesion.Instead, indices capture a wide variety of dimensions such as political freedoms, democracy, and perceptions of economic opportunities.Moreover, indices use a standard set of questions for quantifying levels of trust, which may or may not correspond to the social cohesion issues most salient within a given context.For instance, for a context where out-groups are not a dominant concern, the question on trusting other people may be interpreted differently by respondents.In surveys specific to social cohesion, this could be addressed by asking a set of questions that attempt to capture the same element, allowing for validation at the analysis stage.When using existing indices, one option may be to use these as starting points for additional analysis to allow for contextualization at a project level.
Given the tradeoffs of breadth versus depth, perceptions data have limitations when it comes to their use for monitoring, predictive and evaluative purposes.It is difficult to meaningfully interpret changes in perceptions through statistical methods alone given the intangible nature of such concepts -for instance, more ground level context through qualitative studies may be needed to understand how an increase in trust towards out-groups, as captured in surveys, manifests contextually.
Finally, given their large-scale nature and regular data collection cycles, these indices may be poorly suited to tracking impacts of specific events.While social cohesion overall is a sticky concept that generally changes slowly over time, it is sensitive to shocks that can effect significant changes.To understand the impacts of specific shocks on social cohesion, additional work beyond existing indices may be needed given the importance of timing of data collection.

TRACKING CHANGE S I N BE H A V I O R S
Measurement strategies should ideally track both attitudes as well as actual behaviors.Behaviors that can be tracked include participation in community-based collective action, interactions with other communities or outgroups, and interactions with local formal and informal institutions.Tracking these behaviors over time and across geographical units can yield insights into variation in social cohesion trends across different contexts.Primary surveys can be deployed at regular intervals to track self-reported household or individual behaviors (for example, interactions with local government institutions) or organization reported behaviors (for example, participation in voluntary organizations or village meetings).Incidence of events can also be tracked through primary surveys that collect information on lived experiences of specific events (such as conflict incidents), media reports (such as on incidence of violence in an area) or official and standardized statistics on the same (such as government collected crime or violence data).

EXISTING SOURCES
Questions on tracked behaviors are included in barometer surveys, as well as the World Values Survey and Gallup.This includes questions on actual participation in voluntary associations or community groups or incidence and frequency of interactions with local institutions like the police or frontline government officials.
As with the indices discussed above, given that these surveys ask about both attitudes and behaviors, and are collected at regular intervals of times and across regions, they form a key resource for understanding social cohesion, including temporal and cross-country comparison.
Beyond surveys, there are several existing databases that track incidents of violence and conflict, ownership of arms and other proxies of conflict.These include the Armed Conflict Location and Event Data (ACLED) dataset, which collects dates, actors, locations, fatalities and types of a broad range of conflict events, largely by drawing on local news.Several countries also maintain databases to track local violence, including the Bangsamoro Conflict Monitoring system in the Philippines which collects violence data in Mindanao, or a national level system in Indonesia.While these incidents are hard to tie down at an individual level, they provide rich insight into the experience of violence at a community level.

ADVANTAGE S
Data sources that track behaviors are key in validating perception surveys and can be useful in designing and assessing prevention strategies.Behavior-related data helps assess the extent to which perceptions do or do not align with the attitudes of respondents.For instance, respondents may largely claim to trust their neighbors, but not actively participate in neighborhood associations, which can help identify disconnects and potential entry points.While these datasets can be very useful, it is important to understand their limits.For example, on behaviors, these instruments do not track the very diverse manifestations of participation and their variation across different population groups.Similarly, on incidents of violence or conflict, there may be under-reporting of conflict in remote areas or other areas with limited media presence -and trendlines may be difficult to interpret if constraints on the media change over time.Finally, the impact of violence is broader than its numeric quantity -the quality of a violent incident in terms of the nature, location or target of the attack may carry meaning for local dynamics beyond its quantity.

SURVEY EX P E R I M E N T S
Survey experiments refer to questionnaire design techniques that help indirectly elicit responses regarding sensitive issues during primary data collection through surveys.List and endorsement experiments are two widely used survey experiments. 10  List experiments present respondents with a list of items, experiences, or statements, including sensitive items, requiring respondents to report how many items in the list pertain to them.The researcher then divides the sample randomly into two groups -the direct response group who are presented with a list of neutral and nonsensitive items and the veiled response group, who are presented with an identical list plus a sensitive item.
Respondents are asked to state the number of items in the list they agree with (but never asked directly whether they agree with the sensitive statement).Researchers then estimate the proportion of people to whom the sensitive item pertains by subtracting average responses between the groups.Figure 5 illustrates the structure of a typical embedded list experiment.In the example below, if respondents in the direct response group agree with two items on average, and respondents in the veiled response group with 2.5 items, we can deduce that about 50 percent of the population agrees with the sensitive statement. 10Self-administered field surveys where researchers allow respondents to self-administer responses using electronic devices as an alternative to more complicated experimental approaches is an emerging technique to solicit sensitive information as well.An experiment in rural Philippines tested the effect of this method in response rates and falsification of data and found that allowing respondents to enumerate these questions themselves versus audibly stating their choice reduced nonresponse rates but did not significantly impact falsification rates (Haims, 2020).
Endorsement experiments rely on subtle cues, where respondents rate their support for policies endorsed by socially sensitive actors.With this method too, random variation must be introduced in the data collection strategy.Typically, respondents in the control group are asked to rate how much they support a specific issue or policy, while respondents in the treatment group are asked the same question, but the policy is presented as endorsed by a specific category of actors.With a large enough sample, a difference in the responses of the treatment and control sample can help estimate the support for the specific group in question.Figure 6 illustrates an embedded endorsement experiment.Moreover, survey experiments can be used in combination to validate findings.A 2014 study found that, when carefully designed and analyzed, two survey experiments can produce substantively similar empirical findings even in challenging research environments like Afghanistan and be used to validate each other (Blair et al. 2014).Specifically, the authors uncovered similar patterns of support for the International Security Assistance Force among Pashtun respondents, using both the list and experiment methods in assessing attitudes towards competitors of the state.
The two key challenges associated with this tool are statistical and logistical.Statistically, while indirect questioning techniques reduce bias, they also elicit less information and result in inefficient estimates.This can be partially remedied by combining measurement experiments, e.g. using different survey experiment methods to produce more precise estimates.Logistically, these experiments can be difficult to administer, requiring efforts on the part of the respondent to consider a wide set of factors at once and thus leaving room for misunderstandings and resulting errors in capturing accurate responses.Moreover, simply because they are participating in a survey, even with indirect questions respondents may still perceive sensitive issues to be open to breaches of confidentiality, leading to potential desirability bias.

BEHAVIORAL G A M E S
Behavioral lab or field games are an emerging tool used to complement surveys in understanding interpersonal and intercommunity relationships.Drawing from the field of behavioral economics, voluntary games are a tool used to measure and compare the preferences of the players, particularly their willingness to cooperate with one another.The choices that participants make in a game setting in terms of prosocial behaviors such as trust, cooperation, willingness to sacrifice for the public good are presumed to reflect choices that they would be willing to make in the real-world situations.

ADVANTAGE S
Given their nature, behavioral games can be useful in validating perception and observational data and can be specifically useful in understanding the impact of social cohesion interventions on willingness to engage in collective action (bonding) and attitude towards outgroups (bridging).For example, Alan et al. ( 2021) used this method to investigate the impact of a perspective taking curriculum on the incidence of violence and interethnic friendships in elementary schools in Türkiye, where one in five students was a refugee.Implemented over a one-year period and involving three-hour weekly lectures, activities, and videos, they tracked changes in behaviors using field games testing trust, reciprocity, cooperation, and altruism among participants of the study.
Similarly, Avdeenko and Gilligan (2015) investigated the impact of a community development fund in four states in South Sudan using lab games to measure prosocial behaviors.Importantly, both projects also collected perception and observational data to assess against the results of these field games, to help build robust conclusions.For instance, the study in South Sudan showed no impacts on prosocial behaviors through games but showed an improvement in perceptions of civic participation among participants, allowing the authors to distinguish between the impact on individual norms versus institutions.

CONSTRAINTS
Behavioral games need to be assessed for their limitations in reflecting real world conditions, and extrapolations from games to real world behaviors need to be done carefully and calibrated within local context, including perception and observational data.While games can be iterated and localized, they remain bounded in time and in a simplified context, in a way that does not symmetrically mirror real life considerations.Moreover, in any replication of real-life contexts, timing, location, and setting dynamics need to be carefully considered -for instance deciding when to play games in the context of when the intervention is introduced or where the games are played within a community and how that may affect results of the exercise.

RANDOMIZED CONTROLLE D TRIALS
Randomized controlled trials are designed to be used for evaluation and monitoring purposes and investigate identification or attribution issues.Random variation can be introduced at different levels, including at the individual or community level, or other units of intervention used in a project.Randomized experimental approaches have been widely used in studying elements of social cohesion in various contexts and are increasingly being used even in challenging conflict-affected settings.Given that randomized evaluations collect data at regular intervals, metrics tracked at various intervals can be used for monitoring as well as evaluation purposes.
Depending on an intervention's theory of change or the specific research question, this approach can be adapted to evaluate and monitor effects on all three dimensions of bonding, bridging, or linking, including in conflict settings.For example, Blair et al. (2019)

ADVANTAGE S
Randomized evaluations are an effective tool to demonstrate impact, especially of programmatic components.They help parse out the influence of multiple coexisting factors and impacts of specific variables.For example, an evaluation can test different combinations of inputs such as cash transfers and dispute resolution training and their impact against individually offering these in similar settings, allowing practitioners to have confidence in their specific iteration of the program.

CONSTRAINTS
The biggest constraint to randomization is feasibility, which can be constrained by several factors.For example, randomizing the allocation of cash transfers can be politically difficult, especially in conflict zones where the environment may be unstable to begin with.Randomization is also the most useful when looking at standardized interventions, which can be reasonably administered similarly across implementation areas, helping draw conclusions on the effectiveness of the intervention conceptually and not on implementation challenges in administering the intervention.RCTs also do not capture mechanisms unless explicitly designed for, making them hard to use in settings where the mechanisms are yet to be explored.Moreover, while their relatively longer time to implement means changes in sticky concepts like social cohesion should be captured adequately, attribution becomes difficult depending on the timing of measuring the treatment effect.

FOCU S G R O U P S
Focus groups are a long-standing data collection method that have been used effectively in research on social identity (Posner, 2005) and trust (Paluck and Green, 2009), especially to provide researchers with means to generate meaningful, locally contextualized hypotheses.Focus groups are often used to complement other research methods and involve discussions between small groups of people, where the researcher introduces open-ended questions which participants then discuss among themselves.
Memory workshops are a particular sub-type of focus groups meant to create a collective narrative on challenges communities have faced.They are particularly useful to track the history of formal and informal local institutions and the history of local feuds between communities and out-groups.

Advantages
Focus groups can provide a deep, localized understanding of social dynamics and can be useful to explore the degree of consensus on issues facing a community and the mechanisms driving changes in social cohesion.They are inherently social in nature and data is generated through conversations and interactions that are in line with local practices and only affected by the researcher in limited ways (Cyr, 2017), thus providing insight into "collective sense making".They provide information that can be analyzed at three levels -the individual's, appropriate for triangulation, interactions between two individuals, appropriate for exploration and within the group, appropriate for pretesting (Cyr, 2015).The understanding of mechanisms can help form hypotheses on the determinants and nature of social cohesion that can then be further validated with other data.
Memory workshops can generate rich data on the stories and lived experience of communities, especially when used in conflict settings.In Rebelocracy, Arjona (2016) draws from memory workshops conducted between 2004 and 2012 in Colombia to understand social order during the civil war by examining the motivations of rebel groups and their interactions with existing governance structures.Yaylaci (2020) uses the cases of the Kurdish insurgency in Türkiye and Maoist insurgency in Peru as examples of focus groups, used in complementarity with personal narratives, helping researchers culturally anchor themselves and access reliable accounts of the past by unearthing stories of wartime events and dynamics.

Constraints
The structural benefits of focus groups can also be a weakness in certain settings.Given that focus groups are typically small and formed through non-probability sampling, it can be difficult to select a representative group of participants.Conversely, selecting a representative group can obstruct free-flowing conversation and increase social desirability bias.Moreover, while focus groups can be useful to observe specific group dynamics, these dynamics may also threaten the validity of the data collected -for instance, in a group that comprises the elites and marginalized members of a community together, the actual dynamics between the groups may restrict or give too much voice to the opinions of socially dominant groups or even fall prey to "groupthink", where the opinion of the whole group subdues more nuanced individual opinions.In addition, focus groups are often asked to consider hypothetical situations, which may lead to more socially appropriate discussions of settings rather than actual actions that would play out.
A number of strategies exist to maximize the utility of focus groups.A good practice is to consider data from a variety of combinations of different groups, and to use other data to inform the composition of groups.
Researchers should endeavor to create open and permissive environments, where opinions are welcome, along with possibly asking respondents to record their individual responses before sharing with the group to helping understand potential divergences between individual and group responses.
In addition to the general challenges in the use of focus groups, memory workshops have the added difficulty of researcher and participants grappling with contested and unreliable memories, which can be tainted by time, official state discourses and current ideologies (Yaylaci, 2020).Beyond questions of reliability, the use of memory workshops also involves ethical questions, as they can surface memories of difficult and possibly violent pasts and require facilitators equipped to manage and navigate emotions in the situation to avoid retraumatizing participants.

KEY I N F O R M A N T I N T E R V I E W S
Key informant interviews are personal, semi-structured interviews that are open-ended enough to allow respondents to structure their personal narratives on a topic.Using guiding questions, the researcher can delve into different dimensions of social cohesion without interrupting the respondent's personal stories.Given the nature of the data, this tool is suitable for understanding social cohesion issues more than monitoring, predictive or evaluative purposes.

ADVANTAGE S
Personal interviews with target groups are an important source of data to complement group-based data collection, particularly on sensitive questions, as it allows research participants to contribute anonymously.Moreover, they allow respondents to express diverging or controversial opinions in a safe setting if sufficient rapport is built with the interviewer.

CONSTRAINTS
Data collected from key informant interviews must be assessed for reliability and quality.In terms of data reliability, it can be difficult to conduct interviews with a representative sample, similar to the challenges of putting together focus groups.Personal narratives need to be viewed in the context of influence of identity, political views, and other impacts.Similar to memory workshops, it is important to consider the impact of describing difficult or traumatic experiences on a respondent.

PARTICIPATORY A P P R O A C H E S
Participatory or ethnographic approaches, commonly employed in sociology and anthropology, rely on participant observation, with the goal of generating an immersive understanding of community-level dynamics.Increasingly used in conflict and peacebuilding research, they can provide information of the formal and informal processes that exist in a community for decision-making and can thus be useful to examine how collective action unfolds and what mechanisms and institutions communities rely on.Researchers pay attention to specific questions but also immerse themselves in context and understand people's narratives beyond interviews and focus groups in the context of their daily lives.

ADVANTAGE S
Participatory approaches are useful for monitoring and even predictive purposes and can be a useful complement to evaluations.They can specifically be useful in observing subtle impacts of project interventions in the daily lives of beneficiaries, for example through observing the sessions where beneficiaries are trained in alternative dispute resolution and observing subsequent community meetings, providing rich implementation data.They can open researchers to what they should be monitoring in terms of indicators and help in the observation or recognition of adverse effects of the intervention.

CONSTRAINTS
Ethnographic evidence needs to be considered within the context of both observer bias, where the researcher's priors shape the understanding or collection of data, as well as observer expectancy effects, where participants modulate their behaviors because they are being observed.Moreover, these types of studies tend to require significant time and resources, making them difficult to maintain and replicate at scale.

MAPPING O F C O N S T I T U E N C I E S /ACTORS
The visual mapping of key constituencies, actors and formal and informal institutions is meant to capture the overlapping sets of institutions and actors that drive attitudes and behaviors related to cooperation and collective actions, and that enforce norms.

ADVANTAGE S
Mapping exercises are particularly useful to understand drivers of local grievances and patterns of competition between actors.For practitioners, they can be visually helpful in identifying and taking stock of context at present and over time, more than as a measurement tool for diagnosis, monitoring, or evaluation.Mapping can be done through key informant interviews or focus groups and other available information.

CONSTRAINTS
There is a tradeoff between capturing complexity and maintaining visual and user focused simplification.This can be addressed by being clear about the audience and purpose of the map, and curating information accordingly.

IDENTIFYING PRIORITIES ALONG THE MEASUREMENT CONTINUUM
As the above section shows, there are a broad range of approaches to understanding social cohesion.In identifying an appropriate measurement strategy, it is thus useful to begin with a clear definition of the main purpose of data collection along the measurement continuum: will data be used to inform project and intervention design, to monitor implementation of an existing project, provide early warning tools, or to evaluate impact?Being clear about the objective helps to identify specific data needs, including indicators and collection frequency.Mapping this against an assessment of what data is already available begins to frame the needs for new data collection, which in turn can be assessed against feasibility and advantages and disadvantages described above to settle on the most appropriate data collection tools.

DIAGNOSTICS
A strategy for social cohesion diagnostics relying on multiple sources of data can help ensure accurate contextualization, particularly in conflict-affected environments settings.Qualitative tools in particular, such as focus groups and key informant interviews can shed light on local dynamics, but only when conducted by experienced teams who can carry out such work while minimizing bias.Multiple data sources are often useful to validate initial assessments, for instance comparing behavior and incidents data with findings from surveys focused on people's perception.At the diagnostic stage, the understanding of possible adverse impacts and spillovers of similar interventions can be useful in refining the project design and monitoring structures.Visual tools like stakeholder mapping can be useful in helping project teams process available information.

MONITORING
In collecting data for monitoring project interventions, timing and availability are key constraints.Project design will inform the most relevant indicators, and the team must consider whether available data sources collect relevant indicators frequently enough to fulfill monitoring needs.If surveys need to be put in place, the team must evaluate if the data being collected captures both short-term expected and unexpected outcomes, to allow them to build useful feedback systems within the project.Moving beyond typical quantitative indicators, the team could consider whether the project and context offer space for ethnographic research streams where immediate impacts of implementation can be observed in daily contexts over time, including to identify potential adverse impacts.If monitoring requires the collection of sensitive information, such as support for rebel groups, surveys could utilize emerging tools such as list and endorsement experiments to elicit sensitive information anonymously.

ANTICIPATO R Y A N A L Y T I C S O R EARLY WARNING SYSTE M S
In designing early warning systems, the challenge is in accurately identifying relevant indicators that can predict conflict trajectories.This requires strong diagnostics, including inputs from local stakeholders through qualitative approaches.Once indicators have been identified, teams will need to assess the existing availability of such data versus collection needs.This can involve integrating data from multiple partners and will require validation approaches.Using multiple sources of data can corroborate trends in key predictive indicators: for instance, tracking multiple indicators of trust or incidents of violence can be useful in reaching operational conclusions.Finally, providing space for the emergence and observation of unexpected outcomes in predicting conflict is key to ensure the team is not too narrowly focused.

EVALU A T I O N S B A S E D O N R A N D O M I Z E D C O N T R O L LE D T R I A L S
In setting up a measurement strategy to evaluate an intervention, it is often useful to begin by assessing if randomization is possible technically, administratively, and politically.Evaluations strategy must anticipate and adapt to the political and ethical implications of randomization in different settings -for instance a cash transfer during active conflict versus a dispute resolution mechanism during a post-conflict period -must be considered.
Next, it should be considered whether it is feasible to embed an impact evaluation at the onset of the program itself, which would allow for the collection of accurate baseline information and enable more meaningful comparisons.The quality of implementation, e.g. through process evaluations, is an integral component of impact evaluations given that outcomes are critically dependent on actual implementation.This is all the more relevant for interventions targeting social cohesion, which are context and situation specific along with being heavily dependent on participant behavior, unlike a more simplistic standardized intervention such as the distribution of malaria nets or cash.This process level information of impact evaluations can also be potentially used for predictive or program monitoring purposes.Finally, in the design of the impact evaluation itself, the consideration of adverse effects need to be integrated into the measurement and analysis stage, with space to collect emerging information in dynamic contexts.

MEASURING EFFECTIVELY: CAPTURING ADVERSE AND UNINTENDED EFFECTS
Development interventions can have unintended adverse consequences, which are particularly important in fragile settings and require the practice of conflict sensitivity and application of "do no harm" principles.Barron et al. (2007) examine the specific role of development projects in shaping the nature, extent and trajectory of everyday conflicts and found that development projects in rural Indonesia that gave inadequate attention to dispute resolution mechanisms stimulated local conflict, either through the injection of development resources themselves or by worsening pre-existing tensions in target communities.
Conflict sensitivity involves practitioners considering the context within which their programs operate, understanding the interaction between the intervention and context, and using these insights to avoid negative impacts and maximize positive impacts.It involves considering potential unintended consequences of interventions and taking actions to actively address these."Do no harm" approaches have been operationalized in a number of different tools that apply conflict sensitivity, for example by recognizing the presence of 'dividers' and 'connectors' in conflict and analyzing how an intervention may be implemented in a way that supports local communities address underlying causes of conflict rather than exacerbate them (Haider,2017).Such approaches help increase effectiveness by tailoring interventions to local context, encouraging redressal of unintended consequences and involving local voices and priorities in programming (CDA, 2018).
Measurement strategies should reflect this key principle and be designed to detect both positive and negative effects of interventions on social cohesion.Diagnostics can help in the design of interventions to influence one or more dimensions of social cohesion and can also help identify indicators that can inform practitioners of any unintended negative consequences as the intervention is implemented.The absence of adequate diagnostics and indicators that fail to capture key impacts of an intervention risks exacerbating gaps in social cohesion and creating new challenges.This can include an uninformed distribution of the intervention to those who do not need it and create new grievances or increase the vulnerability of target populations to capture or exploitation.Cillier, Dube, Siddiqi (2018) evaluate the impact of community-level reconciliation forums in Sierra Leone, where victims detailed war atrocities and perpetrators confessed to war crimes.The evaluation showed that while the reconciliation processes led to greater forgiveness of former perpetrators and forged social capital and social networks with people displaying more pro-social behaviors, it also worsened psychological health by increasing anxiety, depression, and post-traumatic stress disorder among participants with both positive and negative impacts persisting for nearly three years after the intervention was completed.
Finally, even projects do not explicitly aim to reinforce social cohesion, it can be nonetheless useful to track social cohesion indicators.For instance, in projects providing cash transfers in conflict settings, it may be useful to track social cohesion metrics over the course of the intervention to observe and detect any increase in intergroup conflicts arising from a particular group receiving monetary assistance instead of another.In Jordan, for example, researchers studied the indirect effects of a cash for work program on social cohesion outcomes, and found that the program strengthened relations between host and refugee communities and offered incentives for more equitable gender roles, in addition to its impact on local economic development (Loewe et al., 2020).In the last years, several studies have provided evidence that interventions can strengthen specific dimensions of social cohesion.This section reviews academic papers that address the impact of interventions targeting at least one of the three dimensions of social cohesion -bonding, bridging, and linking, in both FCV and non FCV contexts.For the section on bridging, the evidence includes a recent systematic literature review carried out by the International Initiative on Impact Evaluation (3ie) on strengthening inter-group dynamics that assessed 24 studies spanning 31 interventions (Doherty et al., 2021), as well as an evidence synthesis published by IPA (Knox et al., 2021).It expands these reviews with the addition of papers published after their conclusion.

REVIEWING THE EVIDENCE ON WHAT WORKS TO PROMOTE SOCIAL COHESION
To do so, we employed the snowball methodology and used Google Scholar to identify papers that cited papers included within the systematic review.Doing so allowed us to use the systematic review as a starting point and then look at additional evidence to build on and assess counter-narratives to the findings of the systematic review along with shedding light on additional interventions not covered in the review that also target dimensions of social cohesion.For the section on linking, in the absence of any relevant systematic review, we used Google Scholar to identify studies that included specific interventions and relevant outcomes related to citizen-state relations.The evidence reviewed here includes study tracking the social cohesion impacts of local economic development programs.This enriches the evidence considered in the 3ie systematic review (Doherty et al., 2021) which restricted inclusion to programs that had at least a component with an explicit social cohesion objective.The 3ie review includes four studies that evaluate the bundling of economic support programs with interventions that encourage contact and dialogue with out-groups.Broader local development programs have typically not been extensively studied to understand their impacts on social cohesion on account of measurement challenges.
In this review, 17 out of the 52 studies included shed light on the effects such programs can independently have on bonding, bridging, and linking cohesion.This helps practitioners better understand the interrelated and even unintended impacts of economic development programming and social cohesion.It can help offer insights on improving existing program design to positively influence social cohesion, identify promising new entry points through economic support programs, and allow for a more integrated and cross-sectoral development approach in resource constrained settings.Some of the studies set in FCV contexts benefit from measurement strategies that capture intangible information.For instance, using public goods games help capture intangible concepts such as cooperative capacity in South Sudan and Turkey (Avdeenko et al., 2015, Alan et al., 2021) and changing behaviors towards out-groups in Lebanon, Iraq, and Nigeria (Doherty et al., 2021).Similarly, survey experiments such as list and endorsement experiments (Lyall et al, 2019;Grossman et al., 2021) help capture attitudes towards governments or support for rebel groups in contexts such as Afghanistan and rural Colombia, where eliciting such information directly is difficult and can create security risks for respondents.
Some of the research reviewed also charts out unintended adverse effects, providing practitioners with more nuanced insights to inform program design.For instance, recent evidence explores the unintended impacts of social cohesion interventions on communities in Bosnia and Herzegovina (Cleven, 2020) and on individual well-being in Sierra Leone (Cillier et al., 2018).It also reveals heterogenous impacts of alternative dispute resolution based on levels political connectedness in Liberia (Blattman et al., 2021) and based on existing political knowledge in Colombia (Grossman et al., 2021).These insights can inform practitioners as they design programs and adopt strategies to mitigate adverse effects.The following subsections look at the evidence on effectiveness of interventions across the three key dimensions of social cohesion.

BONDING: INTRA-COMMUNITY DYNAMICS
For this evidence review, we identified 13 studies that evaluate efforts to strengthen bonding dimensions of social cohesion.Three of these evaluated community-driven development approaches, three covered economic support programs, three covered dispute resolution mechanisms, and four studied security reform programs.
There is a long-standing debate around the contribution of CDD programs to fostering social cohesion.What the evidence to date suggests is that CDD programs have the potential to increase cohesion, including the potential to build and alter cooperative behavior within communities, but that these effects are not automatic or inherent to CDD approaches and -as with other programming interventions discussed here -require careful design and implementation.In Colombia, a rural CDD program improved cooperation within communities, with cooperative behaviors of program beneficiaries influencing non-beneficiaries (Coleman, 2018).A program in Haiti offers additional insights on these dynamics even if it did not involve investments in communities: farmers who were exposed to a framed public goods game posing a collective action problem were more likely to volunteer in a local canal cleaning project, motivated by increasing social connections between neighbors and shifting expectations about each other's contributions to public goods through the game, helping them develop common norms of behavior (Turiansky, 2021).In post conflict Liberia, a development program was able to alter cooperation levels in communities where the program required men and women to address collective action challenges together and not in communities where women worked independently -suggesting that these programs induce improvements in cooperative capacity by providing the space or trigger to do so (Fearon et al., 2015).Yet at the same time, there are multiple examples of where CDD programs failed to affect such behavior changes.Casey (2018) conducted a review of evaluations of CDD programs on social cohesion and trust, including of programs implemented in Afghanistan, Sierra Leone and Sudan and found no evidence that CDD approaches improved trust within a community, potentially in part because these programs built on existing high levels of cohesion.
Similarly, cash transfer programs or livelihood support programs that integrate mechanisms to improve social cohesion can be effective in fostering cohesion.In Malawi, a study comparing the impact of lump sum cash transfers to that of an intervention combining training and participation in a savings group with similar cash transfers found that the combined interventions laid to significant gains in horizontal social cohesion, attributed to the fact that participating in the savings group improved relations within the community itself (Burchi, 2021).
In Côte d'Ivoire, an evaluation of an integrated package of support through village savings and loans associations, business training and encouraging social cohesion showed increased solidarity in target villages as a result of regular meetings of savings groups, but no broader increases in trust or perceptions of security at the community level (Marguerie, 2019).
Introducing non-violent means of dispute resolutions within communities can be effective in improving intracommunity relations, especially in contexts where local clashes are frequent, and these effects can persist over time.Blattman et al. (2014 and2021) evaluated a mass education initiative in Liberia that promoted alternative dispute resolution through an eight-day program focused on problem resolution, negotiation and socio-emotional skills in communities where violent land disputes were frequent.The program was found to be effective at reducing violent and unresolved disputes in the short term, with persistent drops in violent disputes and a small but statistically significant shift to non-violent norms in the long term.It is important to note that these long-term effects are not observed in all contexts.In conflict-affected western areas of Côte d'Ivoire, a community-based conflict resolution project implemented by the UN helped build social cohesion by reducing ethnic and armed violence within the community when measured immediately after the completion of the project, but these effects disappeared over time, with exogenous institutions failing to be reinforced or become integrated within existing community channels for conflict resolution (Khadka, 2020).
Similarly, training local leaders in interest-based mediation can also improve their ability to effectively resolve disputes.Reardon et al. ( 2021) evaluated a program run by Mercy Corps in North Central Nigeria, a region where inter-communal disputes over and resources are frequent and often escalate.The mediation training effectively improved perceptions of security and significantly lowered the share of community members who had recently experienced a violent incident, demonstrating that working with local leaders to improve their dispute resolution skills can be an effective conflict mitigation strategy.
There have been a range of efforts to improve local security, including through state and non-state institutions.
These have included efforts to improve intra-community relations as a way of improving general security within the community and reducing the risk of conflicts escalating to the use of violence.However, the evidence on the impact of such interventions is limited, with additional work needed to understand how to maximize the potential of such efforts.On the positive side, Dyer (2020) evaluated the impact of improving security for rural Kenyan farmers through the provision of watchmen and found that doing so increased perceived security of farmers, reduced theft, and reduced disputes between neighboring farmers.In Colombia, Blair et al. (2021) evaluated the ComunPaz program which sought to replace rebel governance by harnessing complementarities between state and communal authorities through locally embedded mechanisms for sustaining order independently of the state, typically using social sanctions, to improve security and justice provisions in areas once dominated by the FARC.The program included a four-day module training police officers and communal councils to understand the division of labor between state and communal authorities, identify comparative advantages of these groups in dispute resolution and locate sources of conflict within communities.The program enhanced the quality of local dispute resolution and strengthened coordination between state and communal authorities but did not improve reliance on state or communal authorities -possibly because of the increased ability of participants to resolve conflict through dialogue, thus reducing the need for third party recourse.
Interventions targeting the bonding dimension cohesion can have features that interact poorly with existing marginalization patterns.For instance, community-based reconstruction programs have been effective in reaching conflict-affected areas but can be contentious when they only target ex-combatants or conflict-affected individuals for support (White et al., 2018).In Liberia, an alternative dispute resolution training program improved the perception of property rights but was more likely to do so among those more politically connected (Blattman, 2021).In the Philippines, a community policing program increased trust among citizens and their willingness to engage with officers, though these effects again were more pronounced among people related to officers.Moreover, in villages where officers were highly embedded in family networks, there were higher rates of disputes, with people who had no family or social ties to officers rating officer performance more poorly (Haim, 2021).
As the above examples show, it is important to consider unintended adverse effects of programs on social cohesion, particularly the ways in which they may create new social divisions or exacerbate existing fault lines.For example, while cash transfers have been shown to improve social inclusion and capital for marginalized and vulnerable communities, they can also generate intra-community tensions and feelings of being unfairly excluded among non-beneficiaries, for example as a result of eligibility criteria being perceived as unfair or unclear as evidenced from cases in the Republic of Yemen and Kenya (Pavanello et al., 2016).In Indonesia, an unconditional cash transfer program triggered multifaceted conflicts which were accompanied by social unrest.While the government failed to reduce resultant conflicts, community leaders were able to minimize conflicts through informal channels of redistribution (Sumarto, 2020).

BRIDGING: INTER-COMMUNITY RELATIONS
Collaborative contact was found to improve social cohesion through improving a sense of belonging and tolerance14 but showed mixed evidence on trust.On the whole, the effects of collaborative contact on its own appear to be inconsistent and context-dependent.The studies in the systematic review looked at the impact of radio dramas and found positive impacts on trust but heterogenous impacts on other parameters, with particularly strong effects reported during latent conflict and no impacts on violence or extremism.Evaluations of inter-group dialogue interventions found indications of adverse effects on trust as well as acceptance of diversity. 15he review also documents several instances of adverse effects at the individual and community level in the context of interventions promoting dialogue or perspective sharing.For example, in investigating impacts of school-based peacebuilding programs, Cleven (2020) evaluated the impact of a short-term dialogue project in Bosnia and Herzegovina where activities included seminars for local politicians and municipal administrators, parents, teachers, as well as joint activities for school children.The study found that the intervention actually reduced in-group trust.At an individual level, Cillier, Dube, Siddiqi (2018) examined the effects of community level reconciliation forums in which victims describe violence they experienced and perpetrators ask for forgiveness in post-conflict Sierra Leone.As noted, while the study found positive effects on trust towards excombatants, it also revealed adverse impacts on the psychological health of participants that lasted for a year.These observed adverse effects highlight the need for careful monitoring and mitigation strategies at the design and implementation stage.
We find three additional studies beyond the systematic review that show positive impacts of perspective taking interventions.In Cyprus, an evaluation of an extended contact workshop -where indirect contact entailed learning about the experience of others -those who participated exhibited greater trust in outgroups and support for cross-ethnic interaction, but only among those who were initially more opposed towards reconciliation (Donno et al., 2021).In Colombia, the research team varied the protocol of facilitating discussions among ex-combatants and members of conflict-affected communities and found that those in the treatment involving perspective-giving improved their attitudes towards ex-combatants substantially, especially when compared to those in groups where an argumentative approach was encouraged (Ugarrizo et al., 2016).
In Hungary, participation in an online perspective taking game led to more positive sentiment towards Roma and refugee communities (Simonovits et al., 2018).
Inter-ethnic exposure and interactions in daily life can also have positive impacts on trust.In Israel, randomly assigning Jewish patients to Arab doctors across 21 medical clinics facilitated intergroup contact and improved majority group perceptions of minorities -with contact with Arab doctors reducing the social distance preference of Jewish patients and increasing their perceptions about the feasibility of peace (Weiss, 2020).In Türkiye, an early childhood program preparing Turkish and Syrian children for primary school led to the formation of interethnic friendships, improved the Turkish language skills of Syrian children (especially in classes with more Turkish children), and had a long-term effect on reducing primary school absenteeism for Syrian children (Boucher et al., 2020).
At the same time, the benefits of exposure can be erased by key contextual elements.For example, in Türkiye, in classrooms where a program successfully strengthened ties between Turkish and Syrian students, teachers' ethnic prejudice expressed through biased classroom practices significantly lowered the prevalence of interethnic social links, increased within-group ties among host children, and put refugee children at a higher risk of peer violence (Alan et al., 2021).Mass media campaigns can work, with impact depending on the messenger and their social importance to the communities to whom the campaign is targeted.However, the evidence on impact is mixed and points to the importance of paying attention to unintended effects of the message being promoted.Blair, Littman, Nugent (2021) assessed the impact of radio messages in Maiduguri, Nigeria (the geographical origin of the extremist group Boko Haram) delivered by religious leaders emphasizing the importance of forgiveness, announcing the leader's forgiveness of repentant fighters, and calling on followers to forgive.The study found that radio messages delivered by trusted authorities led to large and positive changes in people's willingness to accept Boko Haram fighters back home, and make people think their neighbors are in favor of reintegration.On the other hand, in Burkina Faso, pro-peace religious messaging targeted at adolescents in school was found to have no effect on attitudes towards religious extremism, and instead had the unintended effect of increasing intolerance towards ethnic others by inadvertently reinforcing ethnic identities and reminding respondents of how their community had been the target of religious violence (Grossman, 2021).

LINKING: CITIZEN-STATE RELATIONS OR VERTICAL COHESION
Strong citizen-state relationships, meaning in particular that communities view local state institutions as legitimate, trustworthy, and reliable, are key in preventing conflict and in stabilizing post-conflict societies.A diverse set of interventions, ranging from economic support to citizen engagement to improved service delivery and security and justice reforms all aim to shape how citizens and communities perceive the state.We identify 29 studies covering interventions influencing citizen-state relations through various interventions.These include economic support at the community and household level, social accountability measures integrated into government programs, information provision, and security and justice reforms.
Both community and household level economic support have the potential to reduce incidence of violence, improve perceptions of the state and reduce support for rebel groups and other non-state actors.These can be effective through a variety of mechanisms -improving perceptions of service delivery and governance, improving perceptions of leaders or government officials, improving security in communities, or increasing cost in terms of livelihood opportunities.
The provision of community reconstruction grants or provision of infrastructure help generate physical and tangible evidence that governments are investing in communities and can have far-reaching effects on how communities perceive governments.In areas of limited statehood, a state's legitimacy among the domestic population crucially depends on whether that population feels safe and secure and security perceptions of the population thus play a key role in strengthening state legitimacy at both the community and county level (Nomikos, 2021).In Afghanistan, the National Solidarity Program improved attitudes towards government and perceptions of security, without affecting security incidents themselves (Beath et al., 2012).In Iraq, violent incidents reduced with community level spending on reconstruction efforts (Berman et al., 2011).A more recent study in Iraq found that political events that raise popular expectations of future public service and security provision increases support for government and decreases support for violent opposition groups (Mikulaschek, 2020).
Household-level economic support programs like cash transfers, public works programs and livelihood training can create better and more sustainable economic opportunities for households and build support for the state.India's public works program, NREGA was found to encourage people to help the police track down insurgents (Khanna and Zimmerman, 2014).The Pantawid Pamilyang program in the Philippines reduced the number of conflict-related incidents and decreased the influence of insurgents (Crost et al., 2016).In Afghanistan, while cash transfers alone increased short-term but not long-term support for government, a combination of cash transfers and livelihood training resulted in lasting decreased support for rebels (Lyal, 2019).In Colombia, the Familia en Accion program was found to have positive effects on the demobilization of combatants, observed three years after the program started (Pena, 2017).Economic support can also improve trust in governing institutions through improvements in perceptions of both leaders and governance.In rural Tanzania, cash transfers increased trust in leaders and perceptions of leaders' responsiveness and honesty, with communities where more information was shared on development projects and revenue use seeing larger improvements (Evans et al., 2019).
However, initial positive impacts of community and household-level programs can risk repercussions.Mitigating such risks is important for program design.For example, in the case of NREGA in India, in conflictaffected parts of West Bengal, civilians helping the police were more vulnerable to retaliation (Khanna et al., 2014).Moreover, when program design excludes specific community members or groups, intervention can undermined vertical cohesion among excluded groups.In Pakistan's Benazir Income Support program, an increase in trust in government and leaders was observed only when relative deprivation in communities was salient, with differences between beneficiaries and non-beneficiaries driving the increase (Kosec et al., 2018) and a significant increase in trust being observed in better-connected communities (Ghorpade, 2019).An evaluation of Peru's Juntos cash transfer program found that the program increased trust in local government institutions linked to the program, but only among beneficiaries of the program.Moreover, perceptions of unfairness in targeting and implementation lowered trust in institutions, a finding emerging in other studies on cash transfer programs as well (Camacho, 2014).
Another consideration is the potential for reduced trust in government in contexts where conflict escalation affects the operation of existing economic support programs.For example, in South Sudan, the intensification of violence in 2016 halted the operations of an unconditional cash grant and life skills training program for the youth, where those who were scheduled to but did not receive the grant experienced strong reductions in trust in government (Muller, 2019).
Integrating social accountability mechanisms within community economic support programs can improve citizen-state relations, as long as mechanisms are effective in getting beneficiaries to report issues and the state acting on this information (Fox, 2015).In Uganda, training beneficiaries of a community program to report misuse of funds led to improved monitoring of funds, project quality, and trust towards the central government (Premand et al., 2017).In Indonesia's Raskin rice distribution program, providing information to beneficiaries in the form of cards documenting eligibility led to a large increase in subsidies received by eligible households (Olken et al., 2018).
Information provision combined with adequate service delivery can improve trust in governments when sustained over time.In conflict zones in the Philippines, sustained engagement of the state through organizing regular meetings with community leaders increased the response rate on COVID-19 cases, enabling increased service delivery to at-risk communities (Haim et al., 2021).The Liberian government's door-to-door canvassing campaign to voluntarily comply with disease containment policies during the 2014-15 Ebola epidemic was effective in improving safety related behaviors, generating support for public health policies and increasing trust in governments (Blair et al., 2020).In Pakistan, a government agency used social media during the COVID-19 pandemic to communicate with citizens and created positive perceptions of management of the COVID-19 crisis, increasing trust in government (Mansoor, 2021).
Information provision can help shift support from non-state actors to the state when combined with demonstrated improvements in service delivery.In Pakistan, providing accurate information about reduced delays in state courts led citizens to report higher likelihoods of using them and demonstrating greater support for the state, as measured in lab games.Moreover, this intervention led to reduced support for non-state actors (Acemoglu et al., 2020).Providing information on service delivery, however, can also reduce trust in governments by increasing expectations.In Colombia, providing information on service delivery through text messages led to a reduction in satisfaction with services, critically among those, with low political knowledge and information, backfiring among those such initiatives typically seek to engage (Nussio et al., 2019).
The evidence on the impact of community policing on trust in state institution is mixed.Community policing to date shows limited success in improving trust in governments and police institutions, possibly attributable to poor implementation of existing interventions.A large-scale systematic evaluation of community policing programs across six countries (Brazil, Colombia, Liberia, Pakistan, Philippines, and Uganda) found that community engagement did not increase trust in police, improve citizen cooperation with police or reduce crime levels.However, the results were attributed to implementation challenges resulting from a lack of sustained buy-in from police leadership, frequent rotations of leaders and officers and a lack of resources to effectively respond to issues raised by citizens.
However, community policing has shown some promise in improving civilian awareness of the law, improving relations between police and communities, providing alternative channels of dispute resolution, and reducing support for rebel groups.In rural Uganda, a community policing program increased interactions between civilians and the police and improved knowledge of the criminal justice system among civilians (Blair et al., 2019).In rural Liberia, police visits within communities facilitated relationship building which in turn improved perceptions of the state (Karim, 2020).In Liberia, the Confidence Patrols program (which trained and better equipped police officers) had positive impacts on knowledge of the police and law, increased security of property rights and reduced the incidence of crimes like assault and domestic violence.Moreover, the increase in crime reporting was concentrated among those disadvantaged in traditional mechanisms of dispute resolution, showing promise of an alternative channel of improving social cohesion (Blair et al., 2019).In rural Colombia, the ComunPaz16 program reduced citizen trust and reliance on rebel groups in previously FARC dominated areas (Blattman et al., 2021).
Lastly, while community policing can foster reliance on the state, this can happen through trust in authorities but also through a fear of the state.In South Africa, increased police visits improved a willingness to rely on the police and promoted less reliance on vigilante groups, although this increase was driven by a fear of state punishment for vigilante violence, rather than an increase in the belief that state capacity had improved (Wilkes 2021).

AREAS FOR FUTURE EXPLORATION
There is limited evidence to date on what works in contexts where the scope of FCV risks continues to evolve and escalate.For instance, one study in the review looked at the impact of a livelihood intervention that had to be discontinued in South Sudan as security deteriorated (Muller, 2019).There is scope to improve our understanding of how programs can change in the face of these evolving situations and what that means for resultant impacts.Even when specific interventions have not been shown to be effective yet, it is important to examine intermediate outcomes and to consider the contextual factors that may have undermined the effectiveness of these interventions.For instance, while systematic reviews suggest that interventions such as workshops facilitating intergroup dialogue (Doherty et al., 2021) or community police interventions (Haim et al., 2021) do not improve bridging and linking cohesion respectively, it is important to consider the ways in which specific modalities of implementation or context-specific factors.In this example, it would be useful to understand if there is a specific style or context within which inter-dialogue workshops can positively influence intergroup relations or if such programs could lead to increased conversations between groups or improve attitudes towards out-groups, if not behaviors.Such nuanced information can help practitioners not only tweak existing programs but also design more effective ones in the future.
More research is needed on the interplay between the intended interventions to influence social cohesion and their actual implementation.Social cohesion interventions often involve multiple components and depend significantly on situational factors that shape program delivery.For instance, the effectiveness of an education program promoting peacebuilding could depend on who is facilitating the sessions, what mode it is delivered in, how student's questions and concerns are addressed or the content itself.It is less straightforward to track the implementation of such programs compared to more transactional programs such as providing bed-nets or textbooks in a school.This can make it difficult to decisively ascertain whether the intervention does not work as an approach, or whether the problem lies with its implementation, underlining the importance of integrating process evaluations within more quantitative evaluations.

CONCLUSION
Social cohesion, e.g. the ability of people to collaborate for the public good, can be an important driver of resilience in contexts affected by fragility, conflict and violence.Both horizontal cohesion (relationships within and among communities) and vertical cohesion (relationships between communities and the state) strengthen the ability of communities to withstand shocks and the ability of governments to deliver services effectively.
Cohesive societies are better able to deal with change, and to resist the influence of armed actors.While social cohesion depends on norms and behaviors that can be deep rooted, the evidence shows that it is possible to strengthen cohesion through development interventions.The above sections have described a broad range of projects that have done so, across different regions, income levels, and institutional capability levels, and addressing both vertical and horizontal cohesion.While the interventions vary in their focus and scope, successful interventions all share an emphasis on careful diagnostics to identify specific constraints and entry points, thoughtful implementation with a focus on adaptive learning and remaining aware of potential adverse effects, and rigorous evaluation to understand impacts and their sustainability over time.
The review of the evidence also makes clear that there are many ways to foster cohesion.This includes fostering cohesion through stand-alone projects (like programs that specifically aim to increase trust in the police, increase inter-ethnic connections in schools, or promote non-violent conflict resolution), but also fostering cohesion as part of programs that have broader objectives (such as programs aimed at increasing livelihood opportunities or improving access to public services).Understanding the potential impact that a broad range of projects can have on social cohesion opens the door to thinking creatively about sectoral interventions and how to maximize their impact in this regard -often through relatively low-cost adaptations in design and implementation.
At the same time, the paper highlights that results were neither automatic nor universal.Instead, the broad range of cases examined paper highlights the importance of: (i) Sound diagnostics, including to identify specific constraints that a project seeks to solve and tailoring interventions accordingly.As the paper shows using the case of the national police in Liberia, perceptions can be driven both by changes in the performance, but also in how they relate to communities; (ii) Quality implementation, including addressing concerns such as perceived favoritism in targeting or an inability to follow through on project promises, which can erase the positive impacts of a project and highlights the importance of effective communication and reliable follow through, and (iii) Careful monitoring and evaluation including to enable the early identification of unanticipated adverse consequences, for example when improvements in irrigation in Nepal reduced incentives for inter-community collaboration, or when participation in a government program makes communities targets for reprisals.
Just as the work on social cohesion spans numerous academic disciplines, its operational understandings are evolving across a community of practitioners, including national governments, international institutions, development agencies and civil society.Finding ways to share and harness this collective knowledge will be critical to making progress on this undertaking.

ANNEX: LITERATURE REVIEW METHODOLOGY
The literature review included studies that addressed at least one dimension of social cohesion, e.g.bonding, bridging, or linking.While studies were not excluded based on research methodologies, a majority of the studies included were designed as randomized controlled trials.
The studies for the review were assembled using the snowball methodology, which involves using a key publication on the subject as a base to search for relevant publications by identifying the papers cited by the document (backward snowballing) and papers citing the document (forward snowballing).This process is then carried forward as more papers are identified and their references are assessed for inclusion.
For studies on bridging interventions, the International Initiative on Impact Evaluation (3ie)'s systematic review "Strengthening intergroup social cohesion in fragile situations" (Doherty et al., 2021) was used as the starting point.This review included 24 studies spanning 31 interventions.The 24 studies included in the review were then used as a base to implement the snowball method and identify 11 additional studies.
For studies on bonding interventions, an evidence synthesis published by IPA and JPAL, "Governance, Crime, And Conflict Initiative Evidence Wrap-Up" (Knox et al., 2021) was used to identify key studies on bonding mechanisms.These studies covered interventions in the areas of community policing, justice provision, peacebuilding and reconciliation and cash transfer programs.These studies were used as a base to identify 14 additional relevant studies for inclusion.
For studies on linking interventions, the 3ie systematic review (Doherty et al., 2021) was used as a base to identify some studies on citizen-state relations along with using Google Scholar.Google Scholar was used to search for studies investigating the impact of economic support at the household and community level, social accountability measures, information provision from the state and security and justice reforms that looked at outcomes such as perceptions of the state, interactions with local government, service delivery and violence indicators.This process yielded 29 relevant studies.

Figure 2 :
Figure 2: Diagrammatic representation of how social cohesion relates to social capital, social resilience

Figure
Figure 3: Measurement Tools and Their Use Across the Measurement Continuum Tool/Measurement Purpose Context Setting

Figure 5 :
Figure 5: Example of List experiment technique in a survey

Figure 6 :
Figure 6: Example of an embedded endorsement experiment in a survey

Figure 4 -Coverage of commonly used perceptions and attitudes-based data surveys
Berman et al. (2011) violence over time can help in designing and evaluating prevention and engagement strategies at the local level.For example,Berman et al. (2011)use data on violent incidents in Iraq to examine the impact of $2.9 billion in U.S. reconstruction funds allocated through the Commander's Crost et al. (2016)Program (CERP) and related smaller programs.Similarly,Crost et al. (2016)use data on conflict related incidents at the village level to examine the impact of a cash transfer program in four provinces in the Philippines.
Blattman et al. (2021)f recurring patrols by local police in rural communities in Liberia on perceptions of the police among communities.Similarly,Blattman et al. (2021)looked at the impact of an alternative dispute resolution program among community members in three counties in Liberia on the resolution of unresolved and violent land disputes, capturing the intervention's impact on bonds between community members.Finally, in post-conflict Sierra Leone, Cillier et al. (2018) looked at the effect of community level reconciliation forums on the relationships between community members and former combatants, capturing relations with out-groups.
This section reviews the literature presenting the existing evidence on what works -and what does not -to promote social cohesion.This section reviews a broad range of interventions, including mechanisms to strengthen local institutions, to repair inter-group dynamics, to improve perceptions of the government and state legitimacy, and to reduce or prevent the incidence of violence.It aims to provide an up-to-date synthesis of what we have learned, though individual studies and systematic reviews, about the type of interventions that may effectively promote social cohesion.

The breakup of evidence across different interventions and mechanisms
The review included 52 studies published since 2011.Our sample includes 14 studies on interventions focused on bonding, 11 focused on bridging, and 29 studies focused on linking dimensions of social cohesion.The evidence reviewed covered a broad range of previously under-studied mechanisms and specific interventions that target social cohesion.Along with recent interest in applying social cohesion concepts to development interventions, the evidence base on the effectiveness of such interventions has continued to grow.While a few years ago most studies on social cohesion focused on community-based programming, recent years have seen an expansion of studies covering a broad range of social and economic interventions.This has included contexts ranging from forced displacement to the reintegration of ex-combatants, and innovative instruments including perspective taking workshops, reconciliation forums and trainings on alternative dispute resolution mechanisms.
(Sonnenfeld et al., 2021)reviewed interventions that target inter-community relations and either implement them independently or in conjunction with other social programs.We draw on insights from a systematic review of interventions targeting intercommunity relations and additionally identify ten studies that assess similar interventions as well as traditional economic support interventions that also measured impacts on inter-group cohesion.These additional studies largely encompass research published after the systematic review, while others are included for measuring impacts on social cohesion even if fostering cohesion was not their principal objective, such as economic support programs.The International Initiative on Impact Evaluation (3ie) in 2021 published a systematic evidence review on what works to improve inter-group relationships across 24 studies, assessing 31 interventions(Sonnenfeld et al., 2021).The review identified five key types of interventions-school-based peace education interventions; collaborative contact 11 interventions; intergroup dialogue interventions; workshop-based peace education with intergroup contact and economic support; and media for peace interventions.Overall, the review found small but statistically significant positive impacts across intervention types and measures of social cohesion, with programs that accurately identified local bottlenecks to intergroup social cohesion having larger and more positive effects.The review found (i) largely positive impacts in school-based peacebuilding and interventions that bundled inter-group facilitated contact with economic support, (ii) mixed impacts in interventions focused on collaborative contact and media promoting peace, and (iii) negative or adverse effects in interventions promoting controlled inter-group dialogue.School-based peacebuilding 12 was found to improve inter-group social cohesion by inducing more pro-social behaviors, and had small positive effects on trust, willingness to participate and willingness to help, as well as weakly positive effects on acceptance of diversity.The review looked at five studies evaluating bundled interventions combining workshop-based peace education interventions with intergroup contact and economic support and found a positive impact on trust, and imprecise but positive impacts on sense of belonging and willingness to participate.
Valli, Hidrobo, Peterman (2019)hood support interventions can also improve social cohesion between groups because they facilitate intergroup contact and collaboration, even without the inclusion of an explicit training or dialogue component.Valli, Hidrobo, Peterman (2019)looked at social cohesion in refugee-host settings while examining a short-term transfer program targeting Colombian refugees and poor Ecuadorians in urban and peri-urban areas of northern Ecuador and found improvements in social cohesion among Colombian refugees in the hosting community through enhanced personal agency, attitudes accepting diversity, confidence in institutions, and social participation, without any negative impacts among Ecuadorian counterparts.The authors attribute the impacts to the joint targeting of Colombians and Ecuadorians, the interaction between nationalities at monthly nutrition sessions, and the messaging around social inclusion by program implementers.In Jordan, a vocational training program enrolled both Jordanians and Syrian refugees, encouraging contact between the groups.An evaluation found that those enrolled in the program showed less in-group favoritism, increased trust with others and less bias towards those from other nationalities as a result of interactions(Mercy  Corps, 2021).A mixed-methods study evaluating a cash for work program in Jordan including both Jordanian and Syrian participants found trust improving between the two groups as a result of shared experiences while working together -including conversations about each other's interests and values, collaborating over shared work objectives, shared meals and leisure activities after work and critically exchanging skills among each other.