Social Protection and State–Society Relations in Environments of Low and Uneven State Capacity

Grounded in social-contractual ideas about relationships between the governed and those who govern, the provision of social benefits to citizens has historically been predicated on expectations of acquiescence to state authority. However, the rapid expansion of noncontributory social assistance in sub-Saharan Africa, often supported by global donors through technical assistance programs, raises myriad questions about the relationship between social protection and the social contract in fragile and low-capacity contexts. Focusing on sub-Saharan Africa, but drawing on the theoretical and empirical literature on social protection from around the world, this review parses out the redistributive, contractual, and reconstitutive effects of social protection programming on citizen–state relations. We argue that program features—including targeting, conditionality, accountability mechanisms, bureaucratic reach, and the nature and visibility of state–nonstate partnerships—interact dialectically with existing state–society relationships to engender different social contract outcomes for differently situated populations. 151 A nn u. R ev . P ol it. S ci . 2 02 1. 24 :1 51 -1 74 . D ow nl oa de d fr om w w w .a nn ua lr ev ie w s. or g A cc es s pr ov id ed b y 2a 00 :2 3c 5: 74 00 :a 90 1: 4c 35 :2 be a: ef 4e :9 6e e on 0 6/ 04 /2 1. S ee c op yr ig ht f or a pp ro ve d us e.


INTRODUCTION
In 2015, the government of the Central African Republic (CAR) established the Londö public works program, sponsored by the World Bank, to provide social protection through temporary employment for 35,500 vulnerable individuals across the country. Overseen by CAR's Agence d'Exécution de Travaux d'Intérêt Public en Centrafrique (AGETIP-CAF) and housed within the Ministry of the Economy, Planning and Development Cooperation, Londö worked with local government authorities in the country's 71 sous-préfectures (local municipalities) to recruit citizens for two months of manual labor. In addition to providing a much-needed source of income in an otherwise sparse labor market, the program delivered a visible public good: clearing and repairing roads to facilitate access to markets, schools, airports, and nearby towns.
Public works schemes like Londö are increasingly popular in contemporary social protection programming. Social protection programs encompass all public and private initiatives that "provide income or consumption transfers to the poor, protect the vulnerable against livelihood risks, and enhance the social status and rights of the marginalized" (Devereux & Sabates-Wheeler 2004, p. 9). Ranging from cash transfers to more conventional social welfare schemes (e.g., food assistance), social protection has comprised a core pillar of global poverty reduction strategy since the early 2000s. Noncontributory schemes have been particularly popular in aid dependent states, where, rather than emerging from the demands of citizens, social protection has instead been embraced by governments and development partners as a technocratic means to alleviate povertyand sometimes to build state capacity (Carpenter et al. 2012, World Bank 1990. 1 Tracing their roots to social-contractual ideas that states and citizens have rights and responsibilities toward one other (Harvey et al. 2007), the provision or extension of benefits by governments to citizens has historically been predicated on governments' expectations of delegation or acquiescence to state authority. Yet social protection programs are increasingly implemented in contexts where governments have limited infrastructure and reach, and where international partners are heavily involved in program design and delivery. Across much of sub-Saharan Africa, foreign development partners provide more than half of a given country's social assistance spending, and in six of 31 countries examined in a 2018 review (Beegle et al. 2018), they provide the country's entire social safety net. 2 Against the backdrop of a major humanitarian crisis, many of CAR's sous-préfectures have faced decades of insecurity and contested territorial control. In 2016, the lack of an administrative state presence in the far northeastern sous-préfecture of Ouadda-Djalle presented a binding constraint to Londö's implementation, since there was no local administrative presence to support and oversee the program. Following a series of negotiations between local, national, and international stakeholders, an office of the sous-préfecture was ultimately established and staffed to enable the delivery of social protection benefits, despite the area's otherwise extremely fragile state-citizen relationships. Representatives from the central government had only limited administrative presence in the region since the country's independence. 3 While Ouadda-Djalle represents an extreme example, Londö was not unique in transforming the visibility and infrastructure of the state as part of global poverty reduction efforts. Indeed, from highly fragile settings to higher-capacity states, governments exhibit considerable subnational variation in their presence, reach, and visibility. These complex realities necessarily affect the ways in which poverty reduction programs are designed and delivered, sometimes requiring interventions, as in Ouadda-Djalle, to enable effective implementation.
The Londö program lays bare three important questions that have implications far beyond lifting communities out of poverty. How, and through which channels, do externally supported social protection programs shape and construct state-society relations? To what extent do the realities and constraints of weak or uneven state capacity mediate the anticipated effects of social protection programs? And, importantly, how do lessons from existing literature on social protection and the social contract travel to environments where social contracts between central governments and their citizens are fragile, contested, or nonexistent (Bossuroy & Coudouel 2018, Hickey 2011, Leisering 2018, Sabates-Wheeler et al. 2015? In this review, we scrutinize social protection in contexts of uneven state capacity where external assistance has been central to program design and delivery (Awortwi 2018, Hickey et al. 2019b, Ouma & Adésínà 2019, Tessitore 2011 Environments affected by instability or unrest are often where state-society relations are most fraught. They are also home to the majority of the world's poorest populations (Corral et al. 2020) and hence represent priority countries for global poverty reduction efforts. Yet the difficulty of conducting research in low-capacity states has led to checkered empirical knowledge. Leading scholars of social protection in the developing world have often focused on developmental states with greater infrastructural power. Emphasizing continuities between European social welfare policy and social protection in developmental states, this line of scholarship foregrounds the instrumental objectives of governments that provide social benefits in exchange for citizen acquiescence (Adésínà 2009, Hickey 2019b). We follow this scholarship in underscoring the centrality of social protection for reconfiguring tacit exchange relationships between governments and their citizens-even given weak infrastructural capacity. Yet we suggest that an emphasis on continuities between traditional social welfare models and social protection administered through technical assistance programs can obscure the ways in which uneven state capacity and preexisting state-society relations interact with particular program features to mediate desired social contract outcomes.
Importantly, we refute expectations of a linear exchange relationship between the distribution of benefits and citizens' delegation of authority, suggesting instead that program outcomes and state-society relations shape one another in sometimes unexpected ways. In doing so, we bring the literature on the politics of social welfare policy in the developing world into conversation with scholarship on state-society relations (Migdal 2001); governance beyond the state (Risse & Stollenwerk 2018, Schmelzle & Stollenwerk 2018; foreign assistance and citizen attitudes (Findley et al. 2017, Sacks 2012, Winters et al. 2018; and citizen trust in government (Brinkerhoff et al. 2018, Levi & Stoker 2000, Levi et al. 2009). We foreground insights from sub-Saharan Africa, where considerable cross-country and subnational variation in state capacity, and an emphasis on 4 For the purposes of this review, we define state capacity simply as the government's ability to implement a range of policy choices (Besley & Persson 2009, p. 1218. This is, as Prichard & Leonard (2010, p. 658) note, "the tersest definition of state capacity"; it is reminiscent of Mann's (1986, p. 59) concept of infrastructural power, defined as "the institutional capacity of a central state, despotic or not, to penetrate its territories and logistically implement decisions." It encompasses not only the bureaucratic capacity and resources to implement policy but also the capacity to oversee and enforce implementation across the entire territory of the state (Soifer & Hau 2008). noncontributory programs such as cash transfers and public works, offer opportunities to evaluate program choices in divergent settings. For background, however, we draw on a burgeoning literature linking social protection to social rights and citizenship around the world.
We advance two related contributions. First, we highlight three specific channels through which social protection programs in areas of low capacity can extend beyond their poverty reduction objectives to reshape and reconstitute states and societies, both independently and in relation to one another. We identify first-, second-and third-order effects of social protection programming on state-citizen relationships, which we characterize under three rubrics: as redistributive (capturing what governments provide to citizens); contractual (capturing what citizens expect from governments as well as their attendant perception of and delegation to authority); and reconstitutive (capturing how states and citizens see themselves, as well as how they act and behave, independently, and vis-à-vis one another).
Second, we direct much-needed attention to the more technical aspects of program design, decision making, and implementation. While secondary program objectives (including building state capacity) have received some scholarly attention, scholars often neglect the practical constraints and limitations-like those evidenced in Ouadda-Djalle-posed by uneven state capacity. We highlight five salient program features that shape-and are shaped by-existing state-society relations and, in turn, can mediate a range of program outcomes. These are (a) decisions pertaining to targeting and inclusion, (b) program conditionalities and accompanying measures, (c) embedded grievance and accountability mechanisms, (d) administrative penetration and bureaucratic reach, and (e) the nature and visibility of state-nonstate partnerships.
Through these two contributions, we demonstrate that state reach and capacity present realities on the ground that affect how programs are designed and implemented. Those same realities, and the program choices they engender, affect other program outcomes, including the reconfiguration of state-citizen social contracts. These dialectic relationships between program and context can create feedback loops that vary widely within and across countries, presenting challenges for scholars hoping to isolate causal effects.
In the next section, we provide some background on the evolution of contemporary social protection programming in sub-Saharan Africa, with a focus on state-citizen relationships. Subsequently, we introduce these first-, second-, and third-order effects, reviewing what we know-and don't-from existing literature. Last, we discuss the aforementioned five features of social protection programming design and implementation that can engender unanticipated as well as intended social contract outcomes.

SOCIAL PROTECTION PROGRAMMING
A rich body of literature traces the emergence and evolution of the European welfare state and its diffusion throughout the developing world (Goodin & Mitchell 2000;Gough 2014;Gumede 2017;Hickey 2019b;Lavers & Hickey 2016;Mkandawire 2010;Rudra 2007Rudra , 2015Wood & Gough 2006). While earlier scholarship emphasized traditional government-led social insurance models, more recent work addresses the centrality of nonstate partners and development agencies, particularly in lower-capacity states (e.g., Cammett & MacLean 2014, Hickey 2019b, Kpessa-Whyte & Béland 2012, Lavers & Hickey 2016. Nevertheless, scholars examining social protection programs in the developing world have emphasized that, rather than being driven by donor agendas, programs are usually government owned (Rudra 2015), grounded in welfare logics (Goodin & Mitchell 2000, Haggard & Kaufman 2008, Schmitt et al. 2015, and shaped by historical trajectories of welfare regimes (Hickey 2019b).
Indeed, Rudra (2007Rudra ( , 2015, Haggard & Kaufman (2008), Hickey (2019b), and others suggest that despite significant donor involvement in funding, design, and implementation, the influence of development partners has often been overstated. These scholars highlight instead the structures of political incentives that governments in developmental states face to provide social assistance to citizens, as a cost-effective means to cement favor or acquiescence from segments of the population.
Literature on the contractual dimensions of social protection in socialist autocracies has similarly demonstrated that citizens may be willing to accept certain restrictions on civil liberties if governments provide jobs, services, subsidized housing, or health care (e.g., Cook & Dimitrov 2017). Many scholars recognize not only that citizens value social assistance but also that social assistance engenders citizen compliance and is less costly for governments than open repression (Fox 1994). Indeed, ruling factions in many nondemocracies have used the expansion of welfare to ensure the acquiescence of citizens through tacit state-citizen social contracts, while clientelist leaders can force poor or vulnerable voters to sacrifice their political rights and autonomy to access distributive programs. 5 These tacit exchange relationships foreshadow contemporary arguments made by Hickey et al. (2019a, p. 11) in the African political context. These scholars posit that when political elites are unable to retain power through the distribution of rents to elites, they may use social protection policies to win or retain electoral support (see also Carbone 2012;Carbone & Pellegata 2017;Harding & Stasavage 2013;Kitschelt & Wilkinson 2007;Van de Walle 2007). Efforts to appease or assuage potential challengers recall similar trends in Latin America, where competitive elections have been among the key determinants of health, pension, and education reforms, significantly altering the incentive structures facing political elites vis-à-vis welfare policy (Touchton et al. 2017). While incentives may be strongest in the context of competitive elections, ruling factions face incentives to legitimate the political settlement in the eyes of citizens whether they are democratically elected or not (Lavers & Hickey 2016, p. 391).
In 1990, social safety nets became a core pillar of the World Bank's poverty reduction strategy, which targeted countries too administratively weak or resource poor to maintain comprehensive social welfare programs (World Bank 1990). Other development agencies followed suit, emphasizing poverty reduction in the foreign assistance strategies of some of the world's leading economies and increasingly centering the language of social protection as the safety net through which governments afford society's poorest and/or most vulnerable a means of survival (Devereux & Sabates-Wheeler 2004;Rudra 2007Rudra , 2015. Although the term social protection does not necessarily imply an exchange relationship with the state, its usage often derives from social-contractual logics of public obligation, wherein states and citizens have rights and responsibilities toward one another (Harvey et al. 2007).
As part of their social protection work, development partners such as the World Bank, various agencies of the United Nations, foreign assistance programs [e.g., the United States Agency for International Development (USAID), the United Kingdom's Department for International Development (DfID), and the Swedish International Development Agency (SIDA)], and others support a range of models. Examples of their social protection programming include pension coverage; maternity, disability, and work injury compensation; noncontributory social assistance; and labor market programs (Barrientos & Hulme 2009, World Bank 2018a. In sub-Saharan Africa, due to limitations in capacity to formulate, deliver, and evaluate programs, the delivery of social assistance typically involves multiple stakeholders, including community organizations, government agencies, national or foreign nongovernmental organizations (NGOs), and development partners (Nino-Zarazua et al. 2012, p. 172). However, even in countries where social assistance is funded almost entirely by donors, de facto partnerships with governments vary considerably. Hickey and coauthors critique the blanket assumption that transnational actors always pull the strings, highlighting examples of full implementation by government agencies as well as resistance to the agendas promoted by international partners (Hickey et al. 2019b, Lavers 2019. 6 While in some cases, international agencies assume the public face of the program, in others, donors and technocrats are silent partners. In most cases of uneven capacity, however, programs fall between these extremes, representing collaborative efforts between national and international stakeholders. The precise nature of these partnerships can have important consequences for citizens' reception of programs, as well as for attitudes and expectations toward government.

REDISTRIBUTIVE, CONTRACTUAL, AND RECONSTITUTIVE EFFECTS OF SOCIAL PROTECTION PROGRAMMING
In this section, we review the effects of noncontributory social protection programming on existing state-society relations, considering varying levels of state capacity. We identify first-order effects as the direct provision of goods or resources from one actor (e.g., a state) to another (e.g., a beneficiary). These effects are termed redistributive, capturing the direct and purported objectives of social assistance, or what Tessitore (2011, p. 13) calls a "minimalist interpretation." They alter the material relationship between states and citizens.
Second-order effects are the repercussions of social protection benefits for citizens' contractual relationships with the state. These effects include citizen engagement with and participation in government; perceptions of the state's legitimacy; and altered government perceptions of, information about, and attitudes toward the population. 7 Many existing studies capture the ways in which the distribution of material assistance-as well as accompanying educational, awareness raising, or participation-based measures-can help governments win favor from citizens, thereby exacting acquiescence and fostering a sense of democratic citizenship. Meanwhile, beneficiaries may also see social protection programs as an indication that governments are fulfilling their obligations to help citizens realize their socioeconomic rights (Kabeer et al. 2010).
However, increased public participation and engagement in social, political, and economic arenas can also reconstitute individuals as citizens with rights, responsibilities, and duties that extend beyond immediate kinship networks and derive from their position within a broader political community. In this sense, third-order reconstitutive effects capture the ways in which myriad aspects of social protection programming, including cursory or incidental features of program design and implementation, reconfigure both states and citizens, independently and in relation to one another. While contractual effects primarily concern what citizens might expect from their 6 In Zambia, Hickey et al. (2019a) show that programs endorsed by DfID and others could be rolled out on a wide scale only when the government embraced cash transfers in 2016, following earlier resistance. In the same volume, Ulriksen (2019) demonstrates that Tanzania resisted World Bank-initiated cash transfer programs, and Seekings (2019) demonstrates that international organizations played no part in the introduction of oldage pensions in Botswana. 7 See Hickey (2011) for a typology of theoretical approaches to understanding the state's social-contractual obligations to citizens. governments (and vice versa), reconstitutive effects shape how individuals see themselves within a broader political project.
Although these classifications are fluid and interdependent, they nevertheless represent conceptually distinct outcomes that can fundamentally restructure state-society relations in sometimes contradictory ways (Migdal 2001). While some aspects of social protection programming, such as targeting, may erode or enhance trust in government, others, like the creation of unified social registries, might reconfigure expectations of what it means to be embedded within a hierarchical political entity (Kieh & Agbese 2007, Selolwane 2007. A cash transfer program may engender positive attitudes toward local politicians, while other features of the same program may introduce or exacerbate tensions in the community. Even without extensive state involvement, therefore, social protection programs can serve to consolidate and centralize-or disrupt-the exercise of state power. Centralizing effects can be economically expedient and politically stabilizing. However, they can also rupture social cohesion. Furthermore, in states with enduring legacies of violence, the enhanced monitoring capabilities that help build state capacity can sometimes be directed toward authoritarian objectives (Beresford et al. 2018). This conceptual roadmap offers a framework for understanding how different aspects of social protection programming might reconfigure states and societies in dynamic and multidimensional ways, as well as highlighting how these effects can vary depending on preexisting conditions.

Redistributive and Contractual Effects: Lessons from Existing Literature
Existing empirical research has predominantly focused on the aforementioned redistributive and contractual effects of social protection programming on social contract outcomes, wherein the distribution of certain benefits shapes perceptions of state legitimacy, confidence in state institutions, willingness to delegate authority to the state (e.g., Sacks 2012), compliance with governing institutions, and participation in voting and elections. Informally, poverty reduction and the alleviation of associated harms can enhance individual and community participation in social, political, and economic life (Carpenter et al. 2012) and can inspire citizens to acquiesce more readily to government authority. Whereas redistributive effects represent the direct outcomes of social protection programs-designed to safeguard against shocks, address vulnerabilities, or redress extreme inequalities (Chipenda 2019, Harvey et al. 2007, Kabeer et al. 2010, Künnemann & Leonhard 2008)-winning favor and fulfilling states' perceived obligations to their citizens represent a shift in citizen-state contractual relations. Contractual effects concern how citizens respond to their governments and vice versa.
A wealth of empirical research finds support for performance-based legitimacy models of social protection, in which effective state-run service provision engenders citizen satisfaction and can also increase trust in government (Levi et al. 2009). Across 17 African countries, Brinkerhoff et al. (2018) find that well-run programs increase citizen satisfaction, particularly among urbandwelling citizens. Conditional cash transfer programs in Peru increased trust in the institutions perceived to be responsible for the transfers (Camacho 2014), and cash transfers in Tanzania increased trust in elected local officials and perceptions of government responsiveness to citizen concerns (Evans et al. 2018). However, Jones et al. (2016Jones et al. ( , p. 1210) highlight that despite social protection's potential to bolster the legitimacy of the state, citizens evaluate the state based on prior interactions; thus "merely receiving services from the state does not mean people feel they are citizens of the state." This is echoed by McLoughlin (2018, p. 527), who calls attention to the fact that performance-based legitimacy models can unravel when citizens perceive unfair service distribution. Historical variations in state-society relations may also influence whether citizens attribute social assistance programs to the state or to external actors. For example, Plagerson et al.  (2012, p. 974) show that recipients of cash transfers in areas of South Africa that have historically been economically and socially marginalized were less likely to attribute the transfers to the state.
In addition to shaping citizens' perceptions of government, social protection programs can also shape electoral behavior-specifically by garnering support for incumbents or those perceived to be responsible for the programs. Sometimes this can be attributed to increased trust in government, but support can also arise through other channels, for instance through increased electoral turnout or new information. In Honduras, the Family Allowance Program increased the likelihood that a sitting mayor would be reelected (Linos 2013). Whether programmatic policies increase voter support for incumbents, however, is unclear. Scholars of various countries suggest that they can: In Mexico, De La O (2013) argues that early enrollment in cash transfer programs increased voter turnout by mobilizing pro-incumbent voters, although a more recent analysis found no measurable effect on voter support for incumbents (Imai et al. 2020). 8 Survey results from Latin America suggest that recipients of social assistance programs are both more likely to vote and more likely to support incumbent candidates, and that this remains true across incumbents' ideological orientations (Layton & Smith 2015). Similarly, Turkey's Justice and Development Party's distributive housing initiatives bolstered the party's electoral success (Marschall et al. 2016).
Postcolonial and state-building legacies, as well as the subjective nature of exclusion and poverty, may similarly structure how recipients understand and respond to the provision of social assistance . Indeed, McLoughlin (2015, p. 341) and others identify a nonlinear relationship between service delivery and legitimacy-one that is conditioned on citizens' prior expectations, subjective assessments of impartiality and distributive justice, and ease in attributing performance to the state. While cash transfers in Kenya have built a more inclusive citizenship for recipients by providing new avenues for citizens to access the state, in Tanzania they have reinforced communities' existing understanding of social welfare as the dual responsibility of the state and citizens (Porisky 2019). Some social protection programs may even undermine productive citizen-state relations. Social protection programs can disempower citizens by consolidating their reliance on the state, entrenching polarized partisan affiliations, or undermining informal community-based social safety nets (Freeland 2009). While Béland et al. (2018) argue that social policies have been instrumental in building a national identity in Ghana, Ibrahim (2019) finds that cash transfers have had no effect on the social contract even as they build state capacity for service delivery. Further, Ampratwum (2019) finds that politicians in Ghana use cash transfers to reward voters, thus potentially undermining the idea of a social contract for all, even if the transfer might increase trust in and connection to the state for select beneficiaries.
Research is similarly replete with evidence on the positive repercussions of social protection initiatives for other aspects of social and public life. Social protection has been found to bolster educational outcomes, which can have direct and indirect, as well as second-and third-order, effects on attitudes toward governments. In some studies, school meals have led to increased enrollment and educational attainment, which can have downstream effects for citizen engagement, as well as for other social and economic outcomes (Attanasio et al. 2012, Aurino et al. 2019. Cash transfers can improve food consumption and redistribute family expenditures toward child nutrition and education (Attanasio & Mesnard 2006; see also Tiwari et al. 2016), increase school enrollment (Baird et al. 2013), increase female entrepreneurship and household food consumption (Attanasio et al. 2012), increase women's empowerment and financial independence (Bonilla et al. 2017, Patel et al. 2015, 9 and improve psychosocial wellbeing (Attah et al. 2016, Kilburn et al. 2016, Samuels & Stavropoulou 2016. Each of these improvements has implications for economic development more broadly, which can strengthen state capacity; alter citizens' perceptions of the role, duties, and obligations of government; and improve citizens' view of the state. Programs might build community and foster new social relationships, thereby generating social capital (Adato 2000; see also Attanasio et al. 2009;Pavanello et al. 2016). However, targeted programs can also create social stigma (Hochfeld & Plagerson 2011) or strain relationships between recipients and nonrecipients, as we discuss more fully below (MacAuslan & Riemenschneider 2011).
For many of these indirect repercussions, the boundaries between contractual and reconstitutive effects are blurred. Indeed, various nonlinear and more diffuse repercussions of social protection programming can shape and reconfigure citizens' relationship to the idea of the state and their own situated identities within it (Migdal 2001). We turn next to these effects.

Reconstitutive Effects of Social Protection Programming
A general assumption built into many social protection programs in low-capacity states is that improvements in social welfare and service delivery can contribute to state-building outcomes (Carpenter et al. 2012). Services can act as a direct line of contact and accountability between governments and their citizens, which makes the state visible to its citizens, sometimes for the very first time. As Van de Walle & Scott (2011, p. 9, cited in Carpenter et al. 2012) observe, services can make the state "tangible" through quotidian interactions. This goes beyond simply "seeing the state" (Corbridge et al. 2005) and enables citizens to "engage with it and challenge it where it falls short of expectations" (Molyneux et al. 2016(Molyneux et al. , p. 1090Bossuroy & Coudouel 2018). But how does strengthening the image of the state in one sector contribute to wider state-building processes or to generative relationships between citizens and states, wherein the governors and the governed have responsibilities and obligations toward one another?
First, social protection programs can alter a recipient's perceptions of their own identity and self-image. Rather than affecting their attitudes toward government per se, educational programs or other interactions with new state institutions can disembed individuals from kinship and familial structures, placing them in a reconfigured relationship with the state wherein they take on new and more autonomous identities as members of a broader political community, with rights and attendant obligations (Merry 2003, p. 351; see also Béland et al. 2018). Some studies showed that recipients of Brazil's Bolsa Família cash transfer demonstrated an increased sense of belonging and agency following participation. The program's accompanying health and educational provisions purportedly alleviated recipients' shame associated with having inadequate school clothing or having to beg for assistance or informal loans (Hunter & Sugiyama 2014), thereby reconstructing participants' dignity and sense of self. Ghana's conditional cash transfer program, Livelihood Empowerment Against Poverty (LEAP), "made beneficiaries feel like citizens and forged a sense of being part of the state," connecting them to the idea of the state above all else (Oduro 2015, p. 31). 10 The issuance of identification cards "served as a form of 'contract' between the state and beneficiaries," while beneficiary forums provided crucial spaces for information exchange and interaction (Oduro 2015, p. 31).
Women recipients of social protection programs can experience unique reconstitutive effects, particularly when those programs elevate their positional power within their families or as citizens. 11 Where gender has been a focus of research, it has primarily been at the household level, examining the impact of transfers on women's participation in the labor market (Asfaw et al. 2014, Barca et al. 2015 and their decision-making power within the household (de Brauw et al. 2014, Wasilkowska 2012). Nevertheless, programs may enable women to start their own businesses, exercise financial autonomy, or join microcredit systems (Barca et al. 2015). Effects can also derive from formal and informal education that they receive as part of program conditionalities. In several studies in Latin America, financial literacy training and/or programs that provide opportunities for women to interact socially can develop self-confidence and lead to increased public participation and political engagement outcomes (Molyneux & Thomson 2011, Soares & Silva 2010. Such activities have effects not only at the individual and household levels but also at the level of the community, sometimes shifting gender relations more broadly (see Bastagli et al. 2016, Holmes 2019. Cookson (2018), however, in the context of Peru's Juntas program, highlights that program conditionalities can reinforce gender inequalities by requiring women to spend significant time-which could be spent on other tasks-to fulfill formal and informal program requirements. Jones & Holmes (2011) similarly highlight that most cash transfer programs in sub-Saharan Africa are either gender-blind or reinforce women's traditional household gender roles as mothers or caregivers, thereby potentially compromising opportunities for gendered citizenship. While the distinctions between contractual and reconstitutive effects can be blurred, reconstitutive effects move beyond what citizens expect from government to how they see and position themselves within a political project.

FEATURES OF PROGRAM DESIGN
Existing scholarship highlights the important role of domestic political dynamics in shaping not only how programs are received on the ground but also how they are designed and deployed (Hickey et al. 2019b). For example, the Londö program in CAR meant something very different to residents of Ouadda-Djalle, many of whom were interacting with central state institutions for the first time, than to residents of sous-préfectures closer to Bangui, who were accustomed to passing through government checkpoints, interacting with agents of the state, and obtaining government permits to sell goods at market. Yet in addition to these differences, the constraints of Ouadda-Djalle also necessitated the implementation of various measures and accommodations to facilitate the public works. Local leaders, government representatives, the World Bank, MI-NUSCA (the United Nations Multidimensional Integrated Stabilization Mission in the Central African Republic), and other partners had to work together to first establish the office of the souspréfecture and ensure a secure environment for program delivery, with assistance and oversight from MINUSCA, before the program could begin.
Prior relationships with the state vary not only by geographical location but also by community. Members of ethnic, religious, or racial minorities may have starkly different experiences interacting with government agents or institutions than members of privileged or dominant groups. In addition, therefore, to considering how local context may matter for program rollout and reception, program architects must also consider how interventions layer onto earlier state-society relations for differently situated groups. In the sections that follow, we consider the shifting relationship between state capacity, program design, and the realities and constraints of implementation. Specifically, we examine five program features that are shaped by existing citizen-state relations and in turn contribute to the contractual and reconstitutive effects that social protection programs can engender.

Targeting
Most social protection programs in sub-Saharan Africa are not universal, and targeting decisions can have crucial implications for state-society relations. In some cases, participation can carry a social stigma that disempowers vulnerable populations (Hochfeld & Plagerson 2011, Roelen 2019. In others, selection into a program can bifurcate the citizenry (Adato 2000, Mettler & Stonecash 2008) or undermine social cohesion (Oduro 2015). Indeed, targeting and selection decisions frequently lead to grievances arising from real and perceived motivations concerning program inclusion and exclusion. In Mexico, eligible recipients in the Prospera cash transfer program reported experiencing tension and anger from nonrecipients (Adato 2000), leading Adato to assert that targeting decisions can introduce or reinforce social differentiation. Similarly, Cárdenas et al. (2008) observe that social assistance programs in six Latin American countries contributed to negative attitudes toward recipients, entrenching a belief that welfare makes people "work less" than they otherwise would. In Peru, even as the Juntas cash transfer increased trust in implementing institutions among program beneficiaries, it had the adverse effect of decreasing trust among noneligible households. Moreover, when a constituency that has been historically marginalized or persecuted has access to benefits that others do not, program participation can exacerbate dynamics of prejudice and discrimination.
In sub-Saharan Africa, targeting and inclusion have been the subject of considerable debate in the early stages of program design. Economists have mostly approached selection through a cost efficiency lens (e.g., for whom will the program have the greatest impact at the lowest cost). Given potential negative externalities created by program exclusion, a growing body of work highlights the relative benefits of universal welfare programs (Kabeer 2014, Mkandawire 2005, Ravallion 2019). Nevertheless, a variety of other approaches prevail. In many contexts, programs utilize categorical targeting to provide support to underprivileged social groups deemed "deserving of support" by either governments or development partners; these groups often include the elderly, mothers, children, and differently abled persons (Bossuroy & Coudouel 2018, p. 158). The constitutions of some countries, such as Ethiopia, Kenya, Mozambique, Rwanda, Sierra Leone, Uganda, and Zambia, name specific vulnerable groups that should be provided public support (Bossuroy & Coudouel 2018, p. 151). Some social protection programs explicitly target women. Categorical or geographical targeting is often employed in conjunction with poverty prediction targeting, in which a proxy means test (PMT) is used to reach those perceived to be most vulnerable. However, the results of PMT targeting may clash with the community's understanding of poverty, leading to conflicts over justifications for inclusion. In Kenya's Hunger Safety Net Program, communities sometimes rejected the results of the PMT, which were seen as inconsistent with the community's understandings of poverty and vulnerability. While the PMT estimated household consumption, community understandings of poverty emphasized multidimensional vulnerabilities, including lack of access to safe water and housing (Silva-Leander & Merttens 2016).
In areas with low state capacity, where information is scarce, community-based approaches place decisions over inclusion with the community itself, often through public meetings (Coady et al. 2004, p. 61). While such models build on local knowledge of poverty and "deservingness" (Grosh et al. 2008), they may be prone to elite capture, thereby risking the exclusion of the www.annualreviews.org • Social Protection and State-Society Relations most vulnerable members and reinforcing existing conflicts and divisions within the community (Conning & Kevane 2002, Mgemezulu 2008. This can be particularly damaging in divided societies affected by political violence. Elsewhere, public lotteries are used to select beneficiaries, such as in Colombia's Empleo en Accion; the government of Zambia's Supporting Women's Livelihoods; or Londö in CAR. Although lotteries are simple in design and often popular, there is a dearth of empirical data on their effects on communities' perceptions of fairness. Deciding to restrict program benefits to particular subsets of a population assumes that the state has sufficient information about society to make targeting decisions. However, national and subnational variations in state capacity may impact the state's ability to effectively target intended participants (Lavers 2020, Porisky 2020. Low administrative capacity is linked with poor targeting performance (Coady et al. 2004), which may exacerbate social tensions (Pavanello et al. 2016). Importantly, the absence of administrative information on individual income is also central in justifying the use of ordeal mechanisms such as work requirements, described as potentially costefficient ways to target the poorest (Besley & Coate 1992). While the existing empirical literature is divided on whether targeting gains outweigh welfare losses caused by work requirements (Alik-Lagrange & Ravallion 2018), public works remains one of the most common noncontributory schemes in sub-Saharan Africa. These information asymmetries can themselves alter relationships between governments and citizens. For example, in making targeting decisions, program funders may support information gathering. In Chad, a nationwide census was conducted to gather information on national poverty levels, from which proxy means cutoffs for a cash transfer were developed (World Bank 2016).
These decisions over who is eligible for program benefits are hugely consequential. Not only do they create the potential for new grievances but they also often map onto preexisting relationships between dominant and marginalized groups (Teklu & Asefa 1999). Far from being apolitical, the above examples demonstrate that targeting decisions can entrench and reinforce existing insecurities and potentially undermine confidence or trust in the state. These dynamics can work against other projected program benefits in sometimes unanticipated ways.

Conditionality and Accompanying Measures
In some settings, social protection benefits may be conditioned on participation in other activities. These might include labor (e.g., public works or cash-for-work programs); respecting particular laws (e.g., enrolling children in school); vaccinations or healthcare visits; or participation in village meetings, savings groups, or financial literacy classes (Ikiara 2009, Myamba & Ulriksen 2016. In other contexts, conditionalities may demand that participants forgo some behaviors that conflict with desired social contract arrangements, for example, the exchange of food for weapons in demobilization/disarmament programs. Other programs have emphasized the contractual nature of social assistance by requiring both the government and beneficiaries to sign "contracts of coresponsibility" to bolster citizens' use of complementary social services (Fiszbein & Schady 2009) and strengthen the ties between citizens and the state (Beegle et al. 2018). When registered in Cameroon's Social Safety Net, beneficiaries sign moral contracts with the government that stipulate actions that households must take, ranging from participation in community public works to health care and schooling or civic participation (Beegle et al. 2018, p. 71).
Particularly in environments of low state capacity or weak existing infrastructure, conditions may necessitate the installation of new infrastructure or the implementation of accompanying programs to ensure that conditions are met. Enforcing conditionality related to public services requires that services are in place; schools may need to be built and staffed or health centers established (Baba-Ari et al. 2018, Ogamba 2020. Conversations between donors, program architects, and government agencies, therefore, may extend far beyond the transactional distribution of benefits to the broader social and political landscape within which benefits are distributed. They may also significantly increase the presence and visibility of the state in the minds of local inhabitants. The impact of conditionalities for the social contract may depend on the design of the program. Hunter & Sugiyama (2014) argue that the effects of program conditionalities are contingent on the particular make-up of a program's coresponsibilities. Through an examination of Brazil's Bolsa Família, they find that conditionalities are more likely to contribute to citizenship building if the conditions placed on participants match citizens' rights and expectations (e.g., for education or health services); can be feasibly met by participants (e.g., recipients have access to the relevant social resources and services needed to meet conditionalities); and that noncompliance with conditionalities is not punitive (e.g., program officials work with recipients to address challenges to meeting conditionalities).
Conditionalities may increase citizens' access to and use of broader social services, thus building a sense of "belonging and agency" (Hunter & Sugiyama 2014, p. 840). However, conditionalities can also undermine the social contract by making access to citizenship benefits contingent on the fulfillment of program conditions (Cookson 2018; see also Freeland 2009). Such effects once again extend beyond features of the programs themselves; they interact with local specificities and, in particular, preexisting expectations of citizen-state relations. In Peru's Juntas program, Cookson (2018, p. 20) found that while program officials emphasized that the program delivered "rights and citizenship" to rural areas, participants experienced "intense social pressure to be responsible mothers and good community members" (p. 8), and that meeting program conditionalities intensified poverty and social inequalities by placing "unreasonable demands on their time and resources" (p. 8). This deeply gendered reception of the program undermined the very sense of citizenship it espoused.
While conditionalities accompanying program participation can transform the visibility and face of the state by increasing access to or information about public services, they can also engender new challenges, particularly where existing infrastructure does not enable full compliance. Furthermore, in environments where citizenship is contested, or membership of a collective polity is fraught or geographically fragmented, conditionalities can take on vastly different meanings for differently situated participants. Without careful attention to participants' prior relationships and interactions with state institutions, conditionalities can work against the intended objectives of a program's accompanying measures by placing unrealistic, exclusionary, harmful, or unwelcome burdens on individuals and communities.

Accountability Mechanisms
Development partners have emphasized "social accountability" mechanisms within social protection programs, which aim to reposition "participants as active citizens rather than as passive beneficiaries" of social assistance (Molyneux et al. 2016(Molyneux et al. , p. 1093. These mechanisms can create avenues for recipients to engage with and hold the state accountable for the delivery of social protection and, in doing so, provide avenues for meaningful exchange between citizens and the state (Hickey & King 2016). Rather than focusing on the top-down delivery of social protection, accountability mechanisms emphasize social assistance as a right and participants as active stakeholders in programs (Cornwall & Gaventa 2000). Access to independent accountability mechanisms that promote the voice and agency of beneficiaries may serve to strengthen citizenstate relationships by providing avenues not only for beneficiaries to voice complaints but also for www.annualreviews.org • Social Protection and State-Society Relations governments to respond to citizen demands (Hickey & King 2016, Sabates-Wheeler et al. 2015. In practice, accountability mechanisms have taken many forms, including information provision, grievance mechanisms, social audits, and participatory monitoring (Molyneux et al. 2016). India's Mahatma Gandhi National Rural Employment Guarantee Act includes independent "social audits" that provide avenues for participants to "build political capabilities" and "engage with and assert their citizenship" (Sabates-Wheeler et al. 2017, p. 22). Common in sub-Saharan Africa has been the creation of grievance mechanisms that enable recipients of social assistance to voice complaints related to program implementation. For example, Kenya's cash transfer programs utilize beneficiary welfare committees to channel complaints and grievances from the community to program officers. Many programs, including in Kenya and CAR, include a hotline that is monitored at the national level.
The impact of accountability mechanisms on the social contract depends in part on the provision of appropriate spaces for collaborative exchange that enable citizens to provide feedback without fear of reprisal (Molyneux et al. 2016, Sugiyama 2016. Thus, while many social assistance programs include community meetings, these spaces tend to amplify some voices and exclude others, potentially creating or reinforcing privilege or entrenching exclusive citizenship (MacAuslan & Riemenschneider 2011).
Accountability mechanisms are also influenced by existing understandings of citizenship rights. Molyneux et al. (2016) find that in Mozambique, inequality and social exclusion limit citizens' awareness of their rights in relation to the state, and thus while grievance mechanisms provide opportunities for citizens to interact with program officials, citizens are unlikely to voice complaints in practice. Where accountability mechanisms are introduced without careful attention to participants' existing expectations and perceptions of the state's role and responsibilities, or in the absence of trust that complaints will be met with impartiality and due process, they may remain unavailable in practice. Worse still, where they are introduced in contexts characterized by fear, political violence, or social unrest, they may compound feelings of mistrust, or engender real or perceived concerns of reprisal and surveillance, either vis-à-vis the state or other members of the community. As with state penetration, therefore, the efficacy, perception, and usage of accountability mechanisms is likely to be highly uneven, and in some cases may shape retention of social protection programs, inhibit willing participation in them, or undermine perceptions of them.

Bureaucratic Reach
In areas of limited state capacity, social protection programs may be implemented in fairly remote locations, where the state has little prior presence or visibility among inhabitants, or where the authority of the state is challenged by armed groups or other forms of community self-governance. For many social protection programs in such environments, building state capacity comprises a crucial secondary program objective. This may require developing physical, technological, and bureaucratic infrastructure to ensure the smooth deployment and monitoring of the program. It may involve improving decentralized administrative penetration (e.g., through creation of regional antennas or the establishment of subnational government offices in remote locations, as illustrated in the example from Ouadda-Djalle). Or it may entail the registration of localities and citizens for the purpose of social protection targeting, monitoring, or evaluation.
The registration of beneficiaries-or of citizens more broadly-is one channel through which social protection programs contribute to both contractual and reconstitutive effects. Across sub-Saharan Africa, there are growing numbers of social registries for beneficiaries that serve to increase state capacity by centralizing information and increasing the state's visibility among marginalized populations. In many cases, registration requires the possession of a government-issued ID card as proof of identity to access cash transfers. Yet an estimated 45% of eligible people in sub-Saharan Africa lack such documentation, with even lower coverage among women and low-income populations (World Bank 2018b). In Somalia, data legislation, weak infrastructure, and low citizen trust in government have each presented challenges to the creation of a centralized national identity system, which currently depends on a fragmented and unreliable internally displaced persons (IDP) registration system to facilitate humanitarian cash transfers. Even where national access to ID cards is relatively high, there may be immense subnational variation. In Kenya, for example, registration in any of the government cash transfer programs requires potential recipients to have a national ID card, but subnational coverage varies substantially. Even with the introduction of a new integrated system, an estimated 50% of eligible adults in northern Kenya do not have a national ID, and social protection applicants, particularly in border areas, thus face difficulties in demonstrating their eligibility (Dahir 2020, Kenya Hum. Rights Comm. 2019).
The identification requirements of social assistance programs may serve to exacerbate existing inequalities, as people face barriers to proving their identities (Salem 2018). This concern is particularly salient for migrant or displaced populations, or for the poorest and most vulnerable, who more frequently reside in border areas or conflict-affected spaces where access to state infrastructure is difficult (Diallo 2018). Prospective participants who have difficulty producing proof of identity (e.g., those without birth certificates) also face considerable barriers to entry. Furthermore, the normalization of ID cards can amplify fissures between program participants and nonparticipants, especially when registration documents become mainstreamed in determining access to other goods and services.
Social protection benefits may increase registration for ID cards, as potential beneficiaries recognize the value of possessing identification documents that can create a clear link with the state with downstream effects for political participation and access to social services. For example, Osofian (2011) finds that the Hunger Safety Net Program in northern Kenya increased applications for national ID cards, particularly among the elderly, while Kisurulia et al. (2015, p. 322) similarly find that other cash transfer programs in Kenya have had a positive impact on access to registration documents that allow beneficiaries to "feel recognized by the state." Given that many recipients may never have possessed or encountered a government-issued ID card before, such a process can have important ramifications for the social contract, particularly with regard to shaping citizens' relationships with state bureaucracy and shaping their attitudes toward the state as a protector/provider. Indeed, the creation of ID cards for beneficiaries matters for program implementation, for the visibility of the state, and for its capacity to collect information and exert authority. However, such processes can also bolster the state's capacity for monitoring and surveillance, which can have detrimental downstream effects on social contract dynamics, particularly in violent or authoritarian regimes.

Nature and Visibility of State-Nonstate Partnerships
Finally, the interactions of various government agencies with program funders and NGO partners in the implementation of social protection programming at the local level have implications for how citizens view and experience the state. Importantly, the visible presence of different stakeholders in program implementation, and the ways in which this visibility overlays preexisting state-society-NGO relations, is frequently underemphasized in program evaluations. Even where donors have been instrumental in shaping the design of social assistance programming (Brooks 2015), their influence on the day-to-day implementation of programs at the local level may be limited . Irrespective of funding, citizens' direct experiences with the local www.annualreviews.org • Social Protection and State-Society Relations government officials responsible for program implementation shape their trust in government (Evans et al. 2018) and their perceptions of state legitimacy (Ayliffe et al. 2017, Risse & Stollenwerk 2018. Iterative contact with state officials through social assistance programming may also expand citizens' access to other government offices and programs, as well as their confidence and ability to make demands of the state (Golooba-Mutebi & Hickey 2010, Porisky 2019).
However, especially in conflict-prone areas, delivery of social assistance may occur primarily or partly through nonstate actors (Brass 2016, Cammett & MacLean 2014, Carpenter et al. 2012, Gentilini & Omamo 2011. Donor financing may limit citizen participation in the design of social assistance programs (Sabates-Wheeler et al. 2015). Moreover, where programs are run by donors, "there is no contract to which these institutions can be held accountable," and therefore, even where discourse around rights and citizenship is embedded within programs, these are tantamount to "consumer rights" and contingent on the continuation of the program (Hurrell & MacAuslan 2012, p. 263). Beyond the formal avenues through which social assistance is delivered, the visibility of the state, donors, and other nonstate actors in official program discourse and local narratives may be consequential for whether citizens perceive welfare benefits as part of their rights as citizens, or simply as a form of externally provided aid (Plagerson et al. 2012, Sabates-Wheeler et al. 2015. Politicians in Kenya have sought credit for the expansion of social transfer programs through manipulating rhetoric about program origins; politicians in opposition areas are more likely to emphasize the role of donors in funding the programs than politicians in incumbent areas (Porisky 2020). Experimental evidence from Uganda suggests that citizens are more likely to actively express support for social protection policies if they are administered by foreign donors, who are perceived as less susceptible to corruption, clientelism, and elite capture of public revenue (Findley et al. 2017). While access to accurate information about program origins and procedures reduces the likelihood of local political manipulation of program benefits (Kramon 2019) and increases citizens' perceptions of government transparency and accountability (Ayliffe et al. 2017), in states with low visibility and reach, or with legacies of violence, information is mediated through prior levels of trust.

CONCLUSION
Grounded in redistributive logics (Carnes & Mares 2015, Mares & Carnes 2009, Rudra & Tobin 2017, social welfare encapsulates the expectation (and attendant obligation) that states provide a base level of protection to society's most vulnerable members. Building on social-contractual ideas that states and citizens have rights and responsibilities toward one other (Harvey et al. 2007), scholarship on the diffusion of social welfare in the developing world frequently emphasizes the instrumental objectives of governments, wherein the provision of benefits is predicated on expectations of acquiescence to state authority.
However, social protection programs are increasingly implemented in contexts where governments have limited infrastructure and reach, and development partners are heavily involved in program design and delivery. Given predictions that up to two-thirds of the world's extreme poor will live in fragile and conflict-affected countries by 2030, areas of contested and uneven state capacity represent key priorities for global poverty reduction efforts. Yet such contexts are precisely where existing social contracts are most fragile. Donors and implementing partners share a belief that social assistance can help bolster state capacity, as well as engender a reciprocal and mutually accountable relationship between states and citizens, supplementary to poverty reduction objectives. These hopes are borne out in some recent studies, which suggest that donor-driven approaches can complement and strengthen rather than replace or undermine state authority (Brass 2016, Winters et al. 2018).
Yet, donor involvement has also long been criticized for introducing dynamics that have the potential to undermine a country's existing social contract or to reinforce inequalities or dependencies within existing state-citizen relationships (Mkandawire & Soludo 2003). Externally funded programs have sometimes reduced the capacity of states to deliver assistance (Cammett & MacLean 2014), undermined constituents' perceptions of the state's capacity or legitimacy (Sacks 2012), fractured partisan dynamics (Lavers & Hickey 2016), or imposed policy interventions that disrupt existing arrangements or do not align with domestic needs or norms (de Waal 1996, Olivier de Sardan & Oumarou 2018, Ouma & Adésínà 2019.
In this review, we have emphasized the ways in which uneven state capacity and preexisting state-society relations feed into these dynamics, as well as into other program outcomes. In addition to disrupting donor expectations and distorting capacity-building effects, existing conditions can impede the instrumental ambitions of governments. Moreover, program decisions taken in response to practical and logistical constraints can play crucial roles in restructuring relationships between citizens and their governments. These effects extend far beyond a program's redistributiveor even contractual-objectives, sometimes reconfiguring how individuals and communities understand their own subjective identities within a broader political community. The results can be constructive and unifying; however, they can also be disruptive, serving to fracture social cohesion, fragment ideas of citizenship, or create new fissures of access and belonging.
While existing literature thus demonstrates an impressive range of social contract outcomes engendered by aspects of social protection programming, we contend that crucial features of program design, such as targeting and accountability mechanisms, layer onto existing state-society relations to shape how programs are experienced and how social contracts are reconfigured. Moving forward, scholars and practitioners should take seriously the ways in which these features of program design and delivery interact with existing relationships to engender different social contract outcomes for differently situated populations.

DISCLOSURE STATEMENT
S.K.D. served as a temporary, paid consultant for the World Bank in January 2019. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, nor those of the Executive Directors of the World Bank or the governments they represent.