SOCIAL PROTECTION DISCUSSION PAPER No. 2601 | July 2025 Refugees and social assistance in low- and middle-income countries: A review of operational experiences Alfredo Manfredini Böhm Pablo Acosta Jeremy Lebow Emanuela Migliaccio © 2025 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: +1 (202) 473 1000; Internet: www.worldbank.org. This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. RIGHTS AND PERMISSIONS The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: +1 (202) 522 2625; e-mail: pubrights@worldbank.org. Refugees and social assistance in low- and middle-income countries: A review of operational experiences1 Alfredo Manfredini Böhm 2, Pablo Acosta 3, Jeremy Lebow 4, and Emanuela Migliaccio 5 Abstract This paper gathers lessons from major development investments and reviews the broader evidence base on the inclusion of refugees and their host communities in national social protection programs in low- and middle-income countries (L&MICs). L&MICs host the global majority of refugees, who often live below the poverty line and in impoverished host communities with inadequate access to basic services. International institutions such as the World Bank and the European Union have invested several billions of dollars to reverse this situation through social assistance programs. Growing evidence suggests cautious optimism about the results of such interventions. Although L&MICs are rolling out safety nets to refugees and their host communities, these initiatives face unique operational challenges. Policy makers should allocate adequate time and resources to build the capacity of government mechanisms for social assistance delivery in the often-peripheral areas hosting refugees, strengthen collaboration with government and humanitarian refugee agencies, and foster host governments’ buy-in to this critical agenda. JEL Codes: F22, I38, I32. Keywords: Refugees, social protection, social assistance. 1 A previous version of this paper was used as Background Paper for World Bank (2023). We thank the inputs provided by the following World Bank colleagues: Afrah Alawi Al-Ahmadi, Syed Rashed Al-Zayed, Bushra Binte Alam, Ashiq Aziz, Patricia Bernedo, Shrayana Bhattacharya, Sarah Coll Black, Ioana Botea, Paula Cerutti, Hugo Brousset Chaman, Kenichi Nishikawa Chavez, Aissata Coulibaly, Alessandra Heinemann, Janet Heisey, Ayuba Sani Hussein, Suhail Kassim, Angela Guerrero Lopez, Erina Iwami, Ikechi Okorie, Mostafa Amir Sabbih, and Christina Wieser. We are also grateful to Wolfgang Stöjetz (ISDC) and Domenico Tabasso (UNHCR) for their valuable inputs. 2 Independent researcher, a.manfredini.bohm@outlook.com 3 World Bank, pacosta@worldbank.org 4 World Bank, jlebow@worldbank.org 5 World Bank, emigliaccio@worldbank.org 1 Executive summary This paper reviews lessons from major development investments and global evidence on the inclusion of refugees and host communities in national social protection systems across low- and middle-income countries (L&MICs). It first reviews the growing literature on the impact of cash transfers for refugees, focusing on the overlap, synergies, and tensions between social protection and humanitarian assistance. Second, it assesses the implementation progress of World Bank–funded social assistance interventions targeting refugees using qualitative and administrative data. The review of quantitative evidence reveals four critical insights into the impact of social assistance programs on refugee populations. First, the available evidence points to positive, substantial effects of cash transfers on consumption, education, and child health outcomes among beneficiary refugee households. Second, the persistence of the effects of cash transfers on poverty and human development among refugees is fragile. Third, interventions have different effects on hosts and on refugees, and can be more or less effective for hosts relative to refugees depending on context. Fourth, establishing the causal impact of interventions remains difficult in the refugee context as ethics, logistics, and methodology of RCT-based impact evaluations are problematic in acute displacement contexts. The broader cash transfer literature holds potential clues for how to address the limited persistence of impacts. In comparison to simple unconditional cash transfers, more comprehensive “graduating from poverty” programs are better able to sustain impacts over time, although there is a price to pay in terms of cost. Graduation programs for refugees have shown promise in several countries for precisely this type of sustained impact. As for expanding the impact on host communities, the literature concludes that the obvious starting point is to include host communities in programming by default. The experience of the “Graduating for Resilience” program in Uganda is highly instructive. This program offered large cash transfers, savings groups, and agricultural and business training to both refugees and host communities, and the program’s evaluation showed large and sustained gains for both refugees and hosts. The case studies from Cameroon, Ethiopia, and Bangladesh provide practical lessons and highlight the complex challenges of integrating refugees into social safety nets. Across all three contexts, limited institutional capacity, particularly in remote, underserved regions, emerged as a key constraint, slowing implementation and impeding outreach. Government 2 ownership and mandates also proved critical: in Cameroon, the lack of a dedicated refugee agency hindered coordination, while in Ethiopia the refugee agency had little experience with safety net implementation. In Bangladesh, development-focused efforts were initially slowed by government reluctance to prioritize the agenda and administrative delays in implementation through third-party UN arrangements. Ultimately, while leveraging humanitarian actors in Bangladesh helped bridge capacity gaps and exceed targets, institutional silos between refugee and national systems still hindered integration and efficiency. Projects that treated refugees and hosts equally, like Cameroon’s, promoted inclusion but would have required stronger, targeted outreach to improve enrollment rates among refugees. Overall, the case studies underscore the need for early, deliberate planning for refugee inclusion, tailored to political realities, local capacity, and context-specific barriers. They emphasize the importance of involving host communities in transfers and the role of local governments with adequate capacity in ensuring successful and timely implementation. They also highlight the importance of coordination across different institutions when delivering social assistance to both refugees and host communities. Finally, they underscore the need for flexible and adaptive policies that can respond to the changing needs of refugee populations. Based on the quantitative analysis and case studies, several key policy recommendations emerge. While early results offer cautious optimism, implementation remains complex. The capacity of implementing government institutions is a particularly significant constraint. Closely tied to capacity constraints is government ownership and mandate. Effective implementation therefore requires long-term investments to build delivery capacity in underserved areas, strengthen coordination between governments and humanitarian actors, and ensure sustained political commitment from host country authorities. When implementing government-led safety net interventions, it may be necessary to budget extra time and resources for capacity building, especially in remote areas. Additionally, governments may consider enlisting humanitarian agencies’ support to speed up implementation. 3 1. Introduction As of the end of 2024, there were 42.7 million refugees worldwide, of which about 73 percent were hosted in low- and middle-income countries (L&MICs) (UNHCR 2025). On average, 12.2 percent of L&MICs’ populations live under the international extreme poverty line (World Bank 2024b); in high-income countries, this share is 0.7 percent (World Bank 2024a). Typically, food security, poverty, and human development outcomes are worse for refugees in L&MICs than the general population and, often, the local community hosting them. 6 Refugee women and girls, who represent about half of the global refugee population (UNHCR 2025), often face higher poverty levels, lower educational and employment opportunities, and far higher exposure to gender-based violence. 7 Since social protection systems aim to protect people from poverty, they might be expected to also serve refugee populations. Unfortunately, this is not generally the case, particularly in L&MICs, where coverage gaps are large even among local host populations (Kool and Nimeh 2021). Refugees in L&MICs have been protected from poverty and food insecurity largely through humanitarian assistance, often in partnership with host-country governments, but via systems parallel to those used by governments to protect their own citizens from poverty. The assumption underlying this trend is that refugee crises are acute but temporary by nature, and that refugees will soon return to their home country. In practice, this has not been the case: in 2021, the average duration of displacement for refugees was about 13 years (World Bank 2023). Policy makers are gradually trying to reconcile divergent principles, assumptions, and practices in refugee protection. The Global Compact for Refugees (GCR), signed in 2017, commits signatories, inter alia, to delivering humanitarian assistance as much as possible through national systems. 8 Since the GCR, substantial investments of humanitarian and development funds have flowed toward “humanitarian safety nets” and the integration of 6 See, for example, Pape et al. (2018) for Ethiopia; Pape et al. (2021) for Kenya; Voluntas (2021) for Sudan; and Revel (2020) for Türkiye. 7 See Klugman, 2022 and Box 1 in this paper. 8 The GCR’s program of action, point 66, reads “Humanitarian assistance remains needs-driven and based upon the humanitarian principles of humanity, neutrality, impartiality and independence. Wherever possible, it will be delivered in a way that benefits both refugees and host communities. This will include efforts to deliver assistance through local and national service providers where appropriate (including through multipurpose cash assistance), instead of establishing parallel systems for refugees from which host communities do not benefit over time” (UN 2018). 4 refugees in national social protection systems, particularly with financing from the European Union (EU) and the World Bank. Social safety nets are mentioned in the GCR’s program of action only once, in the context of improving nutritional outcomes. The compact does, however, contain a commitment to channeling more humanitarian assistance through host government systems. Since cash is an increasingly prominent vector for humanitarian assistance, including to refugees, the outcome has been a very substantial flow of funding from international development partners to social assistance schemes targeting refugees and host communities. This paper aims to study the effectiveness of these investments seven years after the signing of the GCR. It combines two sources of information for its analysis. First, it reviews the rich and growing literature on the impact of cash transfers for refugees, focusing on the overlap, synergies, and tensions between social protection and humanitarian assistance. Second, it assesses the implementation progress of World Bank–funded social assistance interventions targeting refugees using qualitative and administrative data. This represents the first in- depth analysis of these sources, with a focus on assessing and improving the effectiveness of government-run safety nets serving refugees in L&MICs. Between 2016 and 2024, the World Bank and the European Union invested several billion in safety nets and safety net-like interventions for refugees hosted in L&MICs. A growing body of quantitative evidence on the effectiveness of these interventions indicates benefits as diverse as greater consumption, access to basic services, and improved education and health outcomes—though the impacts may be short lived. Meanwhile, evidence on the performance of these interventions following their implementation by host governments is limited. This paper’s review of World Bank–funded projects suggests that L&MIC governments have made impressive strides toward rolling out safety nets to refugees and their host communities. The model is working, in summary, but these projects face unique operational challenges. The paper concludes by recommending additional efforts to develop the capacity of local safety net delivery mechanisms in the often-peripheral areas hosting refugees; strengthen collaboration with government and humanitarian refugee agencies; and foster host governments’ buy-in and focus to maintain implementation speed. The paper proceeds as follows: Section 2 provides a brief conceptual framework and reviews the experimental and quasi-experimental literature on the effectiveness of cash transfers and safety net interventions for refugees and their host communities. Section 3 reviews qualitative and administrative data from the European Union and the World Bank and key informant interviews on the structure, successes, and challenges of World Bank–funded 5 interventions rolling out safety nets to refugees and hosts. Section 4 concludes by summarizing and proposing suggestions on how to overcome some of the emerging challenges. 2. How have cash transfers benefited refugees and their host communities in L&MICs? 2.1 Social protection for refugee situations in L&MICs Social protection encompasses a wide-ranging set of policies and programs generally aimed at protecting people against poverty and risks to their livelihoods and well-being. 9 The major instruments of social protection can be broadly categorized into three buckets: social assistance, social insurance, and labor market policy measures (see figure 1). Figure 1. Various types of social protection Social protection Labor Social assistance Social insurance market policies Cash In-kind Social transfers transfers services Note: The present paper’s topics are shown in green. Social assistance provides transfers of money, goods, and social services to poor and vulnerable people, without relying on the contributions of beneficiaries or their employers. In L&MICs, such assistance largely overlaps with social safety nets, which protect families from the impact of economic shocks, natural disasters, and other crises, and typically include cash transfers to the poor, public works programs, or the provision of free health care for those unable to afford it. Social insurance provides cash and in-kind benefits but requires contributions from the beneficiaries or their employers. Contributory pension schemes, 9 The categorization and much of what follows is lifted from Loewe and Schüring (2021). 6 employer-contributed paid sickness leave benefits, and health insurance are all examples of social insurance. Labor market policy measures, meanwhile, seek to promote employment and entrepreneurship (active labor market programs) or help maintain the income of those out of work (passive labor market programs). In general, refugees in L&MICs are formally or informally excluded from labor markets (Ginn et al. 2022) 10 and the broader social protection systems (Kool and Nimeh 2021). As refugees are typically among the poorest residents of the countries hosting them, they are generally unable to provide contributions to social insurance schemes, even if they could legally access them. Therefore, this paper focuses on social assistance and, within that category, cash and in-kind transfers. 11 According to UNHCR data, 19 of the top 20 refugee-hosting L&MICs have existing safety net programs, defined as social assistance programs transferring targeted cash or food to the poor. Humanitarian assistance, on the other hand, is generally characterized as assistance “intended to save lives, alleviate suffering and maintain human dignity during and after human-made crises and disasters associated with natural hazards , as well as to prevent and strengthen preparedness for when such situations occur ” (Development Initiatives 2024). Cash transfers and vouchers 12 are among the primary vectors of humanitarian assistance, and account for about 23 percent of international humanitarian assistance budgets. For the sake of brevity, this paper will reserve the term “safety nets” for social assistance programs run by governments, whereas “cash transfers” are those made through both humanitarian and government interventions. Social assistance and humanitarian assistance thus share a focus on relieving the plight of the most vulnerable segments of society. Although the boundary between the two is drawn in different ways across the literature, a key distinction is the time frame and the institutions providing the support. Social assistance is generally part of a social protection system that 10 Ginn et al. (2022) establishes that “refugees face restrictive barriers to their right to work in practice [or] are unable to access lawful employment but have some access to justice for workplace violations,” at best. Notable exceptions are Argentina, Brazil, Bulgaria, Colombia, Costa Rica, Rwanda, and Uganda. 11 Schuettler and Caron (2020) provide a recent review of job market interventions for refugees and internally displaced persons (IDPs). 12 For the sake of brevity, this paper largely avoids the mention of vouchers; reference to humanitarian cash transfers can generally be taken to encompass vouchers, unless otherwise specified. 7 operates long term in support of poor and vulnerable strata of society. 13 Humanitarian assistance delivers short-term, life-saving, and livelihood support to populations affected by conflict or natural disaster. In addition, social assistance is most often government-run, whereas humanitarian assistance is generally led by international bodies, especially in L&MICs (Roelen et al. 2018). However, in the case of refugees, even these distinctions are blurred. As soon and as long as populations are displaced beyond the borders of their home state, they fall under international protection, specifically that of the United Nation’s High Commissioner for Refugees (UNHCR). However, as refugee situations become increasingly protracted, the UNHCR’s humanitarian support stretches into the long term. This in turn blurs the line between humanitarian assistance and social protection. Indeed, the UNHCR’s heretofore prevailing medium- to long-term model of assistance to refugees typically incorporates multiple aspects of social assistance (UNHCR 2007). In that model, refugees receive humanitarian transfers of food or cash and provision of free shelter, health, education, and other basic services, depending on the needs and context—including on whether they are hosted in camps—for years or even decades. As social protection systems continue to develop in L&MICs, and refugee situations become increasingly protracted, it makes sense to explore how these systems could dovetail with or even take over humanitarian support of refugees and their host communities. Several research programs launched in recent years examine how social protection and humanitarian interventions can synergize in refugee situations. Important policy insights can be gleaned from the multiple case studies covered by Oxford Policy Management’s research on the shock responsiveness of social protection systems; the Overseas Development Institute’s research program, “Social protection responses to forced displacement”; the Institute for Development Studies’ Better Assistance in Crises (BASIC) research program; and the Partnership for Economic Inclusion’s research around productive inclusion interventions for refugee programs. Drawing on case studies from the Philippines, Pakistan, Mozambique, Mali, Lesotho, and the Sahel region, Oxford Policy Management’s research explores the potential role for long-term social protection systems in response to large-scale shocks, and opportunities for the coordination and integration of humanitarian interventions, disaster risk management, and 13 An exception was social assistance support during the COVID-19 pandemic. 8 social protection. It argues for strengthening routine social protection systems as a foundation for resilience and suggests enhancing their shock responsiveness through advance planning, data integration, and intersectoral collaboration, while carefully balancing trade-offs such as timeliness, targeting accuracy, and coverage during emergencies (O’Brien et al. 2018). The Overseas Development Institute’s research focuses more squarely on displacement, examining social protection systems for displaced populations in Cameroon, Colombia, and Greece. The case studies (synthesized in Lowe, Cherrier, and Holmes 2022) highlight that the effective inclusion of displaced populations requires adjustments across all phases of the complex process of delivering social assistance, known as “the delivery chain,” following Lindert et al. (2020). Key links of the chain requiring adjustment include outreach, registration, eligibility determination, and benefit provision. For social protection systems to effectively reach the displaced, legal frameworks that explicitly include this population, political will to support them, and adequate funding are crucial to addressing the peculiar barriers that they face, including lack of documentation, cultural disconnection, and the logistical challenges inherent in reaching people on the move. BASIC’s research program aims to answer the question: “In protracted crises, how can international, national, and local actors work together to strengthen commitments and effectively, efficiently, and sustainably provide social assistance to those in need?” Within the larger category of social assistance, the focus is on the themes of politics and financing, climate and livelihood resilience, systems for design and delivery, and inclusion and participation (Sabates-Wheeler et al. 2022). BASIC examines protracted crises around the world, with a particular focus on Lebanon, Niger, Nigeria, and the Republic of Yemen. The program has produced a wealth of papers, including on the structure and performance of various aspects of the social assistance delivery chain in crisis situations, including the national and international political economy, and local governance of these interventions (Lind 2022; Harvey and Mohamed 2022), transfer modalities (Lind, Sabates-Wheeler, and Szyp 2022; Lind et al. 2023), and targeting (Sabates-Wheeler and Szyp 2022). Finally, the Partnership for Economic Inclusion hosted by the World Bank has launched a survey effort to collect information on productive inclusion interventions for refugees and has started to systematize insights (Heisey, Arévalo Sánchez, and Bernagros 2022). Seyfert et al. (2019) have developed an analytical framework that will help in the discussion that follows. The authors analyze the domains in which refugee assistance from humanitarian agencies can be integrated in national social protection mechanisms, and to 9 what degree. This assistance and these mechanisms can have multiple touchpoints along the delivery chain of social assistance to refugees, from financing all the way to monitoring and evaluation. At each point, the degree of integration can vary from null to total. Box 1 summarizes the key terminology. It is important to acknowledge that refugees in L&MICs face different socioeconomic conditions, opportunities, and challenges than do their host communities. For example, refugees are often excluded from formal labor markets and have less access to services such as formal banking. The restrictiveness of policy varies considerably across different L&MICs, and there has long been evidence that inclusive employment and access-to-finance policies can help substantially even out the playing field for refugees (Clemens et al. 2018). In a recent example, a quasi-experimental study examines the socioeconomic impact of the Permiso Especial de Permanencia (PEP), a Colombian government program regularizing the migratory status of 442,462 forced migrants from Venezuela and providing access to social services, safety nets, and the right to work (Ibáñez et al. 2025). The authors found that program beneficiaries witnessed large improvements in well-being, including a 48 percent increase in consumption per capita and an increase in monthly labor income of 22 percent. The importance of policy context notwithstanding, refugees are often culturally distinct from their host communities, differing in language, religion, and social norms, though these cultural gaps can shrink over time as refugees assimilate. In many L&MICs, public-service- delivery infrastructure and effectiveness are limited, and many refugee-hosting communities are geographically and, sometimes politically, peripheral within their own country. Meanwhile, humanitarian systems for the delivery of basic services hold themselves to relatively high standards and have enjoyed comparatively high investment levels. Thus, in some cases refugees may actually receive better services than their hosts. Given the divergence in the baseline socioeconomic conditions of refugees and their host communities, there is no compelling reason to believe ex ante that the impacts recorded for safety net interventions targeting broader populations would be replicated in a straightforward manner for refugees, or even their host communities. 14 Box 1. Four modes of delivering social assistance to refugees 14 There is a wide literature documenting these differences and the impacts that hosting refugee populations has on socioeconomic conditions of hosts. See, for example, von der Goltz et al. (2024). 10 1. Parallel or stand-alone humanitarian response. Under this option, the delivery of social assistance to refugees is stand-alone and thus separate from national systems. 2. Shadow alignment occurs when the assistance, which is a stand-alone response, aligns with existing or future social assistance programs. 3. Piggybacking combines stand-alone response with working through national systems to deliver assistance.a 4. National-system-led implies that refugees are part of the national caseload. Source: Adapted from Seyfert et al. 2019. a. This category is also referred to as “leveraging” (Lowe, Cherrier, and Holmes 2022). Government safety net programs are highly heterogeneous: some focus on simple cash transfers, while others include public works or behavioral interventions, among other measures. Government safety nets tend to operate in contexts and through delivery systems that are different from those of humanitarian cash transfer programs. The general results outlined in the next subsection, which come mostly from impact evaluations of humanitarian cash transfers to refugees, may not automatically extend to government-led social assistance programs that include refugees. This paper hypothesizes that four aspects in particular can be expected to differ: capacity, ownership, cost structures, and both baseline numbers and the outcomes of social integration. Capacity Operations led or implemented by humanitarian or nongovernmental organizations (NGOs) can have different operational capacities than those run by governments. The operational budgets of humanitarian and NGO-led operations are often higher, and staff tend to be better paid and have better capacity than L&MIC government employees. Models that have worked well on a small scale have failed to replicate positive results at scale. 15 This issue is likely to be compounded by the difficulty of designing programs for nonnationals in general and refugees in particular, be it due to language, cultural, legal, or political economy barriers. Refugee camps tend to be close to borders (Fein and Van Den Hoek 2022), and a feature of the political economy and human resources of governments is that border areas tend to attract and retain less capable administrators (Zhou and Shaver 2021). 15 See Bold et al. (2018), although evidence around the big-push graduation model seems to provide a positive exception (Bossuroy et al. 2022). 11 Implementation of refugee and host operations also requires the collaboration of institutions with rather distinct cultures that may not have interacted before. Ownership Besides capacity differences, an L&MIC government may also have different incentives to implement poverty reduction programs for refugees. L&MIC governments at every level may devote more effort and time to fighting poverty among their citizens than to fighting poverty among refugees. Institutions responsible for refugees’ welfare may be different than the principal actors in the implementation of the government’s safety net, creating a lower sense of accountability for refugee outcomes among those actors. On the other hand, governments could also have stronger political motivations than NGOs to fight poverty among hosting nationals and perhaps directly or indirectly also among refugees. Cost structures Although cost-effectiveness is rarely studied in the refugee-facing literature surveyed above, much of the available evidence suggests that cash transfers are far more cost-effective than other typical forms of social assistance, like food, in-kind transfers, or free service delivery. There is reason to believe that costs, both of delivering cash and of administering any accompanying or complementary measures to cash, can be even lower when such measures are implemented by the government. Overhead costs declined, for example, over the lifetime of Türkiye’s Emergency Social Safety Net (ESSN) program, as responsibility for its implementation shifted from UN agencies and humanitarian actors over to the government and local NGO implementers (European Court of Auditors 2024). Integration at baseline and as an outcome Inclusion in national systems can spill over into deeper integration of refugees in the local socioeconomic fabric of the country. Refugees who are included in national safety nets are brought in contact with national institutions and systems, setting the groundwork for their deeper integration into those same systems. Refugee beneficiaries of safety nets usually receive client identity documents (IDs), like those given to national beneficiaries. The ID itself may grant more expansive rights or may trigger better de facto access to employment opportunities or services. Psychosocial impacts may play a role too. Refugees in L&MICs have complex attitudes toward their host communities and host-country governments that can encompass gratitude, solidarity, indifference, suspicion, hostility, and more besides. Refugee beneficiaries of a government-implemented program that also includes host communities may create more connections with their surrounding communities and feel 12 more welcome and freer to build a future. Or, in contrast, they may feel abandoned by the international community and bereft of their hope of being resettled in a third country. The combined effect of this on mental well-being and on socioeconomic outcomes is hard to predict. Without specific knowledge of how these dynamics play out, we cannot assume ex ante that findings from NGO- or humanitarian-implemented programs for refugees would transfer straightforwardly to government-implemented programs for both hosts and refugees. These caveats form an important backdrop to the next section, which reviews the impact of NGO- and humanitarian-implemented programs, to try to anticipate what could be expected of ongoing efforts to include refugees and their hosts in government-run safety nets. 2.2 Experimental and quasi-experimental evidence of social protection for refugees This section seeks to review the available robust evidence on the impact of cash transfers and safety net-like interventions for refugees and refugee-hosting communities. To draw conclusions applicable to refugee situations, the section focuses first on evidence from humanitarian cash transfers in refugee situations, then turns to emerging evidence from government-run refugee-facing cash transfers. For both, the focus is only on experimental or quasi-experimental evidence. The first piece of evidence in the experimental literature comes from a series of studies in Ecuador, where Hidrobo et al. (2014) evaluated a World Food Program (WFP) intervention for Colombian refugees, comparing US$40/month transfers in cash, vouchers, or food to one another and to a control group. All three intervention modalities increased overall consumption during the intervention period by US$12.30–19.01 per month, equivalent to 11–17 percent of monthly consumption expenditure. The size of the effect depends on the modality: food had the largest impacts, but cash was the most cost-effective in enlarging consumption. 16 Notwithstanding, food vouchers were found to be more cost-effective in improving dietary diversity. A follow-up study from the same team noted that cash, food, and vouchers all caused a reduction in intimate partner violence (by about 6 percentage points [pp] or about 20 percent) and male controlling behaviors of female partners (Hidrobo, Peterman, and Heise 2016). The intervention also improved social cohesion, mostly by 16 A similar effect was found in Aker (2017), though without a pure control group to establish the effect of the program independent of modality. 13 improving the sense of agency, confidence in institutions, and social participation among the refugee population, although positive impacts were less clear for the Ecuadorian host community (Valli, Peterman, and Hidrobo 2019). Syrian refugee situations offer further lessons. In Lebanon, Lehmann and Masterson (2014) applied a quasi-experimental regression-discontinuity (RD) design to study the impacts of winter cash transfers to Syrian refugees in the country. Exploiting the fact that a household’s eligibility for the transfer depended on its location above a certain altitude, the study established that despite a significant increase in beneficiary household expenditure on heating assets, the transfers were not sufficient to keep households warm. The transfers, however, helped in reducing households’ tendency to adopt negative mechanisms to cope with income poverty, including child labor (-6 pp); boosted primary school enrollment (+6 pp); and weakly improved social cohesion with the hosting community in the form of mutual help—both given and received. The finding aligns only partially with a later study by De Hoop, Morey, and Seidenfeld (2019). This study used a quasi-experimental geographic RD design to evaluate a “labeled” 17 but unconditional cash transfer program aimed at increasing afternoon-shift school enrollment among Syrian refugees in Lebanon. The study finds no significant effect on enrollment, a finding in which school capacity constraints “may have played a critical role,” but a positive impact on school attendance: children who were enrolled and whose families benefited from the cash transfer spent 20 percent more time at school than the control cohort. Altındağ and O’Connell (2023) also study two humanitarian cash transfers implemented for Syrian refugees in Lebanon by WFP in a quasi-experimental setting: the multipurpose cash (MPC) assistance program and the “food e-card.” The MPC’s monthly support was equivalent to about US$175 per beneficiary household, delivered in cash, while the food e-card provides a voucher for food items worth US$27 per month. Researchers exploit the targeting mechanism of these programs to apply an RD design. Households are ranked by a vulnerability score determined through statistical means and all households above a certain score are included, while those below are excluded. Comparing households just above with those just below this inclusion threshold, the authors find that beneficiaries enjoy substantial increases in consumption levels, child well-being, school enrollment, household 17 A “labeled” cash transfer communicates to beneficiaries that cash transfers are intended to be used for a specific purpose: in this case, defraying the cost of busing younger children to school and compensating families for the forgone income of older children’s labor. 14 food security, and reductions in negative coping strategies. However, effects were found to have dissipated only six months after the end of the program. The twin studies by Salti et al. (2022) and Moussa et al. (2022) provide further evidence on Lebanon’s MPC. 18 The studies employ a similar RD design to slightly different subsamples of the same data set. The former focuses on household-level measures of well-being, including expenditure, education, employment, and shelter; the latter focuses on short- and long-term health and educational outcomes of children of beneficiary households. Salti et al. (2022) find that receiving two 12-month cycles of cash transfers increases household monthly expenditure by US$60 per month (~9 percent above the control group). Benefiting from a single 12-month cycle of cash transfers increases food expenditures and the enrollment of children in school, and decreases child labor in the short term, but not after discontinuation. Participation in the MPC discourages labor force participation but also reduces unemployment. Moussa et al. (2022) confirm positive impacts of the cash transfers on children’s school enrollment and a decreased need for primary health care, among other health and education outcomes. Contrary to Alt ındağ and O’Connell (2023) and Salti et al. (2022), these effects of the MPC are sustained even after the intervention is discontinued, 19 but the effects are substantially larger for beneficiaries that receive two cycles of transfers. In Türkiye, the ESSN (see section 3 for operational details) has also been studied using quasi- experimental methods. Özler et al. (2021) find that the ESSN “caused meaningful increases in [measures of food security] and reduced the use of negative coping strategies, [e.g., selling assets, child labor and the like]”. The effects dissipated over time, as conditions among both beneficiaries and nonbeneficiaries improved. Although identifying the effects of the program is problematic due to the nonrandom nature of assignment and substantial systematic attrition, the evidence suggests that the program had modest, positive benefits, and may have caused substantial spillover benefits for nonbeneficiary households—one of these spillovers being through nonbeneficiary households “transferring” their children to beneficiary households. Aygün et al. (2024) adopt an RD approach to compare outcomes in child labor and school enrollment of families just below and just above the ESSN’s eligibility 18 Several other studies of the Lebanese MPC adopt a similar approach but with smaller data sets. 19 They find that enrollment is substantially higher (7–8 pp) and child labor substantially lower (-3 pp) among all MPC recipients, including for those who received the MPC in only one year and not in the next. An even larger effect still is seen in beneficiaries’ reduced need for primary health care. This does not directly contradict Alt ındağ and O’Connell (2023), as the sample, time frame, and outcome variables differ between the two studies. 15 threshold based on a dependency ratio. The authors find large effects of the ESSN in reducing children’s participation in labor and increasing their school enrollment, most robustly among boys. However, the eligibility conditions of the ESSN make it difficult to attribute all these effects to the ESSN itself. 20 After Hidrobo et al. (2014), one of the few studies of the effects of cash transfers for refugees based on fully experimental evidence is that of Caria et al. (2021). It focuses on three distinct interventions for Syrians in Jordan: a small one-off cash grant (labeled to encourage use in job searches), information, and psychological support to Syrian refugees and Jordanian hosts. The cash grant has a substantial effect on improving employment and earnings that persists four months on; other interventions also have positive effects, but all dissipate by the four-month mark. Importantly, none of these effects is seen among Jordanian jobseekers, who enjoy, however, a slight improvement in well-being from the cash grant alone. For Africa, there is emerging evidence from Uganda’s “Graduating to Resilience” program. This United States Agency for International Development (USAID)-funded graduation project was implemented by a consortium of NGOs in host and refugee communities living in and around the Rwamwanja refugee settlement, which is home to about 85,000 refugees from the Democratic Republic of Congo. The 6,600 client households of the program received US$580–600 in cash transfers over 12 months and a lump-sum cash asset transfer six months after the beginning of the program. They joined savings groups, received agricultural and business training, and participated in weekly individual or group coaching sessions. In the accompanying randomized control trial (RCT), involving 11,000 households across various treatment and control groups, results at the 30-month mark show that participation causes large increases in household income (+32–45 percent, depending on the treatment arm), productive asset holdings (+40–88 percent), per capita consumption (+18–25 percent), subjective well-being, and perceived food security. Both hosts and refugees benefited, although hosts saw greater positive impact across all measures but consumption, for which the effects were similar (Innovations for Poverty Action 2023). The context, of course, matters. Uganda has one of the most progressive policies among L&MICs toward the economic inclusion of refugees (Ginn et al. 2022). Refugees are granted freedom of movement, the right to work, and in some cases, even access to land for farming (UNHCR 20 Household size, which is used as the forcing variable (that is, the variable based on which eligibility for the program is established), is not continuous. This condition can complicate identification of the effect, as discussed, for example, in Imbens and Lemieux (2008). 16 2024). Rwamwanja is not a refugee camp, but a settlement, open to the rest of the surrounding Kamwenge district. These initial results prove that the graduation formula can have substantial impacts on refugee and host communities, at least where the policy environment is welcoming. Initial evidence is also emerging from government-run, World-Bank–funded safety net interventions targeting refugees. The Government of Niger’s Projet d’appui aux Réfugiés et aux Communautés d’Accueil (PARCA) includes a component promoting access to economic opportunities for refugees, internally displaced persons (IDPs), and their host communities. The intervention comprised a US$200 cash grant, life skills, and business training for both agricultural and nonagricultural activities and was targeted at families classified as medium- poor, meaning “able to meet basic consumption needs.” This intervention was not structured as a traditional safety net intervention, as it did not include long-term cash transfers and did not target the poorest in any of the communities targeted. Nevertheless, initial results suggest that the program has improved incomes in the short run (Fernandez et al. 2024). Three months after disbursement, beneficiaries of the program were 4 pp more likely to report increased incomes, and when they did, their income was 17 percent higher than nonbeneficiaries. There is some evidence that this impact is strongest among refugees and IDPs. Some benefits spilled over into the host community: control households in villages where the intervention took place saw their incomes increase, even if they did not participate directly in the program. Interestingly, while measures of trust among refugees and hosts increased, so did the reported incidence of disputes between the displaced and their hosts. The Republic of Congo’s Lisungi project offers another set of emerging insights. Lisungi was the flagship social protection program of the Republic of Congo, launched in 2013 and extended in 2018 to include refugees from the Democratic Republic of Congo and the Central African Republic and their host communities. This extension of the project covered the main cities of Brazzaville and Pointe-Noire, and the Likouala department inhabited by two distinct groupings of hosts: the Bantu ethnolinguistic majority of the Republic of Congo and the diverse, non-Bantu indigenous populations, many of whom traditionally practice a nomadic lifestyle. Beneficiary selection in Lisungi is based on the collection of the country’s social registry data among potential beneficiaries, enabling statistical scoring techniques to identify the poorest in each community for inclusion in the program, after community validation. The project, one of the first World Bank refugee-facing safety net projects, provided quarterly cash transfers to beneficiary households, conditional on children attending school and getting regular health checkups. The value of the transfers was worth 17 between US$17 and US$220, depending on the number of elders and children present in the household. 21 Additionally, a partially separate group would receive training and grants to commence or expand income-generating activities. A quasi-experimental evaluation of both interventions applied matching techniques to elicit the impacts of the program on the refugees hosted in Likouala against a comparable control group, matched on observable variables. Results from the midline, collected approximately one year after support commenced and while it was still ongoing, indicate that the program had increased consumption (per adult equivalent) by an average of 34.6 percent for both hosts and refugees, across treatment modalities—cash transfers and support to income-generating activities. Lisungi’s conditional cash transfers increased consumption most markedly among refugees but also sharply among Bantu households. It is likely the transfers also increased consumption among the indigenous population, but the effect was much less marked and only marginally statistically significant. Income-generating activities also begat consumption increases, albeit at a smaller scale. Lisungi increased asset holdings markedly across all groups and treatment types, although less for indigenous hosts, and prompted host beneficiaries to engage in more hours of work. Finally, measures of social participation and women’s empowerment also increased significantly across groups (Brück et al. forthcoming). Box 2. The differential impacts of displacement on female refugees and the role of social assistance Women and girls account for about half of the refugee population (UNHCR 2025). The Gender Dimensions of Forced Displacement (GDFD) research program gathers gendered evidence on socioeconomic outcomes among displaced populations – both IDPs and refugees. The 17-country synthesis presented in Klugman (2022) warrants the general conclusion that displacement compounds with gender inequalities to the disadvantage of women especially in access to schooling and economic opportunities. Not only are “displaced households […] generally poorer than non-displaced households” but gender is “an important predictive factor in indicators of chronic deprivation, such as school completion”. Both in comparison to female hosts, and to displaced males, displaced women and girls often face heightened poverty rates, have fewer employment opportunities, receive less education, and incur far higher risks of gender-based violence, particularly intimate partner violence. Certain contexts provide counterexamples, however, or counterintuitively highlight the importance of context-specific assessments. There can be several, context-driven reasons for this gender disparity. Displaced women may have lower endowments (e.g. of land), be subject to adverse gender norms and 21 Based on an average exchange rate of 600 Central African Francs CFA to the US dollar. 18 constraints on their mobility due to care responsibilities, more frequently lack identification or legal documentation, or be more constrained in their access to financial services and to formal labor markets. Alongside stronger and better enforced gender- and displacement-inclusive legislation, better access to childcare, more livelihood opportunities, and more gender- disaggregated data collection, cash transfers emerge from Klugman’s (2022) review as an important instrument to protect and empower displaced women. Experimental evidence suggests that cash transfers have caused reductions in intimate partner violence and partner controlling behavior against Colombian refugee women in Ecuador (Hidrobo, Peterman, and Heise 2016). Unconditional cash transfers and (equivalently compensated) short-term employment have also significantly improved women’s “beliefs about their prerogative to make decisions in the household and their intolerance for intimate partner violence” among Rohingya households in Bangladesh (Hussam et al. 2022). However, the effects of working on psychosocial well-being were strongest for men. These findings indicate the importance of intersectional strategies for women’s empowerment in humanitarian contexts and suggest a role for social assistance. To compensate for gender disparities, national social protection systems rolling out to refugee communities should prioritize women as direct beneficiaries of transfers – much as they should when targeting the general population. In designing and implementing cash transfer schemes, special consideration should be given to displaced women’s peculiar constraints and vulnerabilities, based on an understanding of the socio-cultural context and vulnerability patterns. Tailored measures can include providing accessible and safe childcare, ensuring women’s access to information, and incorporating protection services into social assistance programs. Addressing GBV risks and providing support to survivors among displaced women requires an understanding of local dynamics and contextual best practice, and strong linkage of social assistance and humanitarian programming to support services including access to healthcare, contraception, childcare, and psychosocial support. In summary, four common themes emerge from the literature review. First, the available evidence points to positive, substantial effects of cash transfers on consumption, education, and child health outcomes among beneficiary refugee households. This is in line with evidence from cash transfers targeting displaced persons—and not specifically refugees (Doocy and Tappis 2017)—and the more general effects of cash transfers (Bastagli et al. 2016). Second, the persistence of the effects of cash transfers on poverty and human development among refugees is fragile. Of the studies reviewed, only one—the Uganda study—included 19 a long-term (i.e., longer than one year) follow-up. In that case, effects persisted up to the 30-month mark, but in all other cases, several rounds of data collection saw impacts diminish over time. Longer and larger interventions had stronger impacts than shorter and smaller ones. Third, interventions have different effects on hosts and on refugees. Two of the seven distinct interventions reviewed in this section do not target host communities at all. Of the five interventions that do include hosts, four managed to boost outcomes among refugees but had null or much more muted effects on host communities (labor market outcomes in Caria et al. [2021]; social cohesion Valli, Peterman, and Hidrobo [2019]; income in Fernandez et al. [2024]). In Uganda, the situation is reversed, with hosts generally benefiting more from the same graduation program than their refugee counterparts (Innovations for Poverty Action 2023). In the Republic of Congo, the heterogeneity is more related to ethnicity than to nationality, with greater positive impacts seen both among refugees and the Bantu Congolese than among the indigenous Congolese. Fourth, establishing the causal impact of interventions remains difficult in the refugee context. Only five of the studies reviewed use random assignment of beneficiaries to a program. The RCT approach is among the most robust ways in the microeconomics literature to establish the causality of interventions and has been widely used to evaluate social safety nets. Yet the ethics, logistics, and methodology of RCT-based impact evaluations are hardly ideal in acute displacement contexts (Puri et al. 2017; Elice 2021). Withholding potentially life-saving support from a control group is hard to justify ethically, to enforce practically, and even to exploit fully from an empirical standpoint, since spillovers and contamination in fluid displacement contexts can be hard to predict, let alone control—for example, children changing households and the high systematic attrition rates seen in Özler et al. (2021). Given the socioeconomic contexts and challenges faced by refugees, however, we cannot expect experimental evidence of social protection’s positive impacts to translate straightforwardly into results for refugees. And as aid budgets shrink worldwide, careful evidence-based prioritization is indispensable. Despite the challenges, more experimental and more robust quasi-experimental evidence is needed. The broader cash transfer literature holds potential clues for how to address the limited persistence of impacts. Cash transfers, in general, tend to have diminishing impacts over time (Araujo, Bosch, and Schady 2019; Baird, McIntosh, and Özler 2019). In comparison to simple unconditional cash transfers, more comprehensive “big push” or “graduating from poverty” programs are better able to sustain impacts over time (Kondylis and Loeser 2021)— 20 although there is a price to pay in terms of cost-effectiveness. Graduation programs for refugees have shown promise in several countries for precisely this type of sustained impact. The literature review also shows that cash transfer programs for refugees achieve positive impacts across important outcomes but can struggle to sustain these impacts over time, and to extend them to host communities. Both are significant shortcomings, particularly when the process of hosting refugees is as protracted and politically contentious as it has become. This raises the question of how and whether these two limitations can be overcome. As for expanding the impact on host communities, the literature concludes that the obvious starting point is to include host communities in programming by default. This is, however, often hard to do for humanitarian cash transfer interventions, as they are typically strictly targeted at the refugee community, for a variety of prima facie defensible reasons, including agency mandate; division of labor with hosting governments, NGOs, or civil society organizations; and prioritization (or earmarking) of limited funding. The experience of the “Graduating for Resilience” program in Uganda is highly instructive on both accounts. The program’s evaluation shows large impacts on beneficiary households and spillovers in the community that are sustained over time. Its implementation structure also shows that humanitarian cash-plus programs can be extended to host communities, to remarkable effect, although Uganda’s uncommonly refugee-inclusive policy environment may play a part in enabling both results. Another option suggested in the literature is to channel funding through host-country governments, and to roll out cash transfer operations in hosting areas, including both host communities and refugees. This has the added advantage of being a scalable intervention, as governments can and typically have an interest in scaling up poverty reduction programming, provided it benefits citizens as well as refugees. Many safety net operations funded by the World Bank and the European Union attempt precisely this. Evidence on the effects of these operations specifically on refugees and their host communities is beginning to emerge from their respective impact evaluations 22—although this is not the case for all operations. 22 This includes, for example, operations in Niger, the Republic of Congo, and Ethiopia. 21 3. How well have refugees and host communities been integrated into L&MIC safety nets? This section reviews implementation arrangements and progress of major EU and World Bank social protection projects that serve refugees and their host communities. After a general overview, it provides deep dives into projects funded under the refugee-facing funding envelopes of the 18th and 19th round of the World Bank’s International Development Association (IDA), known as the IDA-18 Refugee Sub-Window (henceforth RSW-18) and the IDA-19 and IDA-20 Window for Host Communities and Refugees (henceforth WHR-19 and WHR-20) 23 respectively. An understanding of how past operations have been faring in this space can help shape forthcoming operations and influence their impacts. As the evidence base on the latter interventions is not yet developed and the operations rely on NGO-led implementation, the rest of this section focuses on the EU and World Bank interventions. 4.1 The EU Facility for Refugees in Türkiye and the Emergency Social Safety Net (ESSN) Between January 2015 and March 2016, over 1 million informal migrants crossed the Aegean Sea from Türkiye to the Greek islands (UNHCR 2016). They were mostly Syrian asylum seekers fleeing the civil war in Syria, as well as Afghans and Iraqis. The Facility for Refugees in Türkiye was established in 2016, with an initial tranche of €3 billion and a further €3 billion approved in 2018. More than a third of each tranche is dedicated to social protection interventions, most of them cash transfers. These include the ESSN (€2.1 billion across the two tranches), the Vocational Education and Labor Market Integration program (€236 million), and the Conditional Cash Transfers for Education program (€199 million). The rest of the budget was mostly dedicated to health and education infrastructure and services, refugee protection, and migration management (European Union, DG NEAR 2021). The ESSN and its spinoff, the Complementary Emergency Social Safety Net (C-ESSN), form the largest EU-funded humanitarian operation. Starting in November 2016, and with funding to last until 2025, the ESSN delivers monthly cash transfers to over 1.8 million refugees, most 23 In IDA-21 papers, adaptive social protection for the displaced and enhanced coordination between humanitarian assistance and social protection programming are identified as important outcome areas under the Opportunity element of the People focus area (World Bank 2024c). 22 of them Syrians. The total budget confirmed so far is about €2.65 billion, and as of November 2024, approximately €2.5 billion had been disbursed. The scheme is managed by the Turkish government’s Ministry of Family and Social Services (MoFSS) with the support of several other government agencies. The architecture of the transfer scheme was initially set up with the WFP’s technical support. Over time, the Turkish Red Crescent and the MoFSS took over more implementation responsibilities, initially in partnership with the International Federation of Red Cross and Red Crescent societies (IFRC) under a 2021 EU contract. The IFRC handed over all implementation responsibilities to the Turkish Red Crescent and the MoFSS in December 2023. The Turkish Red Crescent deals with most of the logistics of collecting applications, outreach to and communications with beneficiaries, and handling of complaints. The MoFSS manages the beneficiary database, including applications to the ESSN and house visits to verify beneficiaries. The MoFSS also cochaired the steering committee with the EU’s Directorate-General for European Civil Protection and Humanitarian Aid Operations (DG ECHO) until July 2023, and Directorate- General for Neighbourhood and Enlargement Negotiations (DG NEAR) from 2023 onward. Any registered refugee household can apply, but the system prioritizes large families, families with large shares of dependents, single parents, single women, the elderly, and the disabled, and excludes anyone with registered assets in Türkiye or a work permit (Türk Kızılay, International Federation of the Red Cross, and CashHub 2021). Payment is via a debit card issued to eligible refugees. Since June 2021, the operation split into the standard ESSN and the C-ESSN. The C-ESSN takes on the most vulnerable cases, providing higher cash transfers to those who have the least capacity for self-reliance and who are most likely to require long-term support. The ESSN does not offer an example of refugees’ integration into a national safety net, although the Government of Türkiye is integrally involved in managing it. The program remains parallel to the Turkish social safety net in terms of implementing agencies, payments, and other modalities, although the MoFSS has a stronger role in the C-ESSN. In the terminology of Seyfert et al. (2019), the ESSN constitutes a parallel system, but one that piggybacks off government systems at several and increasingly more points along the delivery chain, including enrollment and information management, as well as overall management. 4.2 The IDA windows: RSW-18, WHR-19, and WHR-20 23 In July 2017, in the aftermath of the GCR, the World Bank’s IDA opened a US$2 billion financing “window,” with dedicated funding to help L&MICs hosting large numbers of refugees. Access to the RSW-18 was provided to eligible countries on 100 percent grant terms for high-debt countries and on split 50-50 between grant and (concessional) credit terms for moderate- and lower-debt countries. In the spirit of international burden sharing, projects in lower-debt countries that target refugees only could be considered for 100 percent grant terms. Between November 2018 and July 2021, all 14 countries that met the eligibility criteria 24 for the window accessed the RSW-18. Twelve of these countries are in Sub-Saharan Africa and two are in South Asia. 25 The focus of programming was refugees’ welfare and socioeconomic inclusion, legal solutions (including freedom of movement and identification), access to basic infrastructure for host and refugees, livelihood support in hosting areas, support of the return of refugees, and strengthened finances for national budgets strained by hosting. The RSW-18 ultimately disbursed US$1.85 billion in financing, 92 percent of the total amount allocated. The subsequent, 19th and 20th, funding cycle for the IDA included a renamed Window for Host Communities and Refugees (WHR-19 and WHR-20). WHR-19 expanded the annual availability of funding slightly 26 and included a specific grant-based window to help host countries deal with the compounding impacts of COVID-19 on the economic burden of hosting refugees. A requirement was added that, to access the window, operations would 24 These criteria are: “(i) the number of UNHCR-registered refugees is at least 25,000 or 0.1 percent of the population; (ii) the country adheres to an adequate framework for the protection of refugees; and (iii) the Government has in place a strategy or plan acceptable to IDA that describes the concrete steps, including possible policy reforms, toward long-term solutions that benefit host communities and refugees” (World Bank 2020). 25 The full list includes Cameroon, Chad, the Republic of Congo, Djibouti, Ethiopia, Niger, Pakistan, Uganda, Bangladesh, Burundi, Burkina Faso, the Democratic Republic of Congo, Mauritania, Rwanda, and as of WHR- 19, South Sudan. 26 To provide surge financing to IDA countries’ COVID-19 response, the 19th cycle of IDA was compressed from the usual three to two fiscal years. Some funding was pushed to the IDA-20, which was brought forward by one year. While the initial WHR-19 allocation was US$2.2 billion over three years, it was reduced to US$1.27 billion over two years in this process, which represents an increase in annual funding availability of 3 percent over the RSW-18 allocation. 24 have to promote refugee-inclusive policy progress in host countries. The 20th IDA funding cycle also featured a similar window (WHR-20), with an allocation worth $2.4 billion 27. Monitoring and evaluation (M&E) data available on RSW-18 and WHR-19 safety net operations were gathered from World Bank projects and systematized to gauge progress. While an encouraging number of projects are active, coverage of the refugee population is low, both at the global and at the national level, and progress is slower in refugee- and host- facing components than in the components facing the general population. To shed light on the reasons for this, three country case studies are presented, based on a review of the literature and interviews with Bank project teams from 2022. The challenges identified are compared across the projects, and operational recommendations are offered, also drawing on more recent developments in the projects as documented in official World Bank M&E reports. The RSW-18 and WHR-19 portfolios implemented social protection interventions in 12 different refugee-hosting countries. 28 All but one were in Sub-Saharan Africa. Cash transfers are the instrument of choice, often pivoting from existing safety net operations for nationals. Social protection interventions for refugees and host communities were featured prominently across all portfolios, accounting for 24 percent of the portfolio in RSW-18 and 23 percent in WHR-19. In total, across the RSW-18 and WHR-19, and 15 projects, about US$796 million were invested in safety net operations. All but one of these 15 operations include at least a cash transfer component: unconditional cash, conditional cash, or cash-for-work. Most include grants as a complement to cash transfers and a provision to include refugee beneficiaries either in the National Social Registry or in the project’s Management Information System. Table 1 summarizes the features of RSW-18 or WHR-19 safety net operations across 12 countries. In most cases, only a selection of features was implemented. 29 27 As of April 2025, US$1.9 billion has been used by projects. 28 As of end-2024, there had been eight RSW-18–funded and five WHR-19–funded projects that had established a safety net operation for refugees and their hosts. Of the WHR-funded ones, two provided additional financing to existing RSW-funded safety net projects. Two further projects contained a substantial safety net component under a broader operation. 29 The Democratic Republic of Congo and Cameroon are the notable exceptions, since their selection process does not target refugees specifically based on status, but rather allows access to project activities irrespective 25 Table 1. Summary of features of RSW-18 and WHR-19 projects that include social protection interventions Inclusio Emergenc n in Infrastructur Uncondition Conditiona y COVID- Accompanyi Accompanyi social e Features al cash l cash 19 cash Public ng measures ng measures registry/ enhancemen implemente Country transfers transfers transfers works - grants - training MIS t d Bangladesh X O O X O O O X 3 Burkina Faso O O O X X O O O 2 Cameroon X O X X X X X O 6 Chad X O X O X X X X 6 Congo, Rep. X X O O X X O X 6 Djibouti O O O O O X X X 3 Congo. Dem. Rep. X X O O X X O X 5 Mauritania X O X O X X X O 5 Niger O O O X X X O X 4 Ethiopia O O O X X X X O 4 Burundi X O X O X X X O 5 South Sudan X O O X X X O O 4 X Included, with a numerical target Legend O Not included Note: MIS = management information system. The projects collectively planned to reach an estimated 1.43 million refugees and 1.81 million host community members with cash transfers. In the process, 450,000 refugees of refugee status or nationality. In the case of the Democratic Republic of Congo, this reflects a progressive refugee policy, that allows refugees virtually equal access to economic opportunities and basic services. In the case of Cameroon, this reflects also the difficulty of identifying refugees, who do not always have proof of refugee status. 26 would be included in national registries or management information systems of projects, ensuring that they could be included in national statistics on social protection and potentially included in future rollouts of programs. The projects also usually extend the safety nets to the refugee-hosting communities, variously defined. If these targets were achieved, safety nets would reach approximately 24 percent of the refugee population hosted across these 12 countries with cash transfers. However, throughout the portfolio, refugees’ enrollment in government-implemented cash transfers has not proceeded as straightforwardly as for host communities. IDA funds became available to these projects at various dates, ranging from January 2019 to April 2022. Setting aside the interventions in Bangladesh—which, as we shall see, are set up differently from the other operations—only about 11 percent of planned refugee beneficiaries had been reached, in contrast with about 58 percent of host communities, as of May 2022. Toward the end of the WHR-19 implementation period, in December 2024, however, the situation was reversed: 79 percent of refugees had been enrolled, as against 72 percent of host communities. Several operations are still active, but in general, the portfolio seems on track to reach the planned number of beneficiaries. On multiple occasions, however, planned timelines proved too ambitious, as the following case studies exemplify. To shed light on the reasons for the slower implementation speed of refugee-facing safety- net operations, the rest of this section examines three country case studies: Bangladesh, Ethiopia, and Cameroon. The countries are chosen to spread across West Africa, East Africa, and Asia, and to reflect different approaches across the spectrum of integration with government systems, as described by Seyfert et al. (2019). The set of operations in Bangladesh exemplifies how humanitarian cash transfers, education, and health services, initially delivered in parallel to national systems, can transition toward greater alignment, and exemplifies how “easy” cash is in comparison to other emergency responses. The safety- net operations in Ethiopia and Cameroon are, by contrast, government-led and at different stages of maturity, with different approaches to targeting and to packaging interventions. 30 The Cameroon Social Safety Nets for Shock Response project: Refugee inclusion by design 30 Where not specified, the sources for the following subsections are in-depth interviews with Bank team leaders of the corresponding operations. Project documentation and administrative data from the World Bank’s Operations and Project Portal were also consulted . 27 Implementation category: Government led Cameroon is a pioneer in the process of integrating refugees in safety nets. The Cameroon Social Safety Net for Shock Response project comprised three cash transfer schemes; once the project received additional financing from the RSW-18, all three were opened to refugees on an equal basis as nationals. Active since 2013, the project laid the first foundations of a cash-based safety net in a country that had lagged behind its neighbors in implementing such a process. It presents an interesting case study because of its ambition to include refugees across a variety of interventions, adopting what we might term an “agnostic” approach to nationality and refugee status: any poor household living in a targeted area was eligible, irrespective of whether host or refugee. Neighbor to several conflict-prone countries, Cameroon hosts a large population of about 490,000 refugees, most hailing from the Central African Republic and Nigeria. 31 Refugees from the Central African Republic have been migrating into the East and Adamoua regions of the country since 2004, fleeing ongoing conflict within the Central African Republic. Starting in 2014, Nigerians have sought asylum in the Far North region of Cameroon, to escape the violence and insecurity caused by the jihadist insurgency of Boko Haram. All three regions were already among the poorest in the country before the refugee influx. The Far North region also witnessed massive internal displacement, as the insurgency staged attacks in the region (World Bank 2018). About three-quarters of refugees from the Central African Republic live in rural areas, integrated with the local population, whereas most Nigerian refugees live in camps managed by the UNHCR. WFP provides social assistance to just about a quarter of refugees, in the form of mixed cash and food assistance, although due to funding shortfalls, the assistance is increasingly sparse and targeted to the most vulnerable (WFP 2025). No specific government agency exists in Cameroon that deals with refugees; instead, various government ministries fulfill distinct functions in refugee policy implementation 32 (Levine et al. 2022). The government launched the IDA-funded Social Safety Net Project (Projet Filets Sociaux, PFS) in 2013 to establish a system targeting poverty reduction and the promotion of human capital development. Until 2013, social protection in Cameroon featured small, ad hoc schemes offering geographically sparse coverage and mostly regressive universal subsidies. 31 Authors’ calculations, based on UNHCR data. 32 For example, the Ministry of Territorial Administration is responsible for refugee management in rural areas, while refugee status determination is handled by the Ministry of Foreign Relations. 28 The PFS aimed to directly benefit 420,000 of Cameroon’s poor with two main interventions— unconditional cash transfers (TMO) and cash-for-work (THIMO)—both of which included accompanying training and lump-sum grants to beneficiaries. A third intervention, emergency cash transfers (TMU), was started in 2016 in response to an influx of refugees from the Central African Republic. Additional financing in 2019, including a grant from the IDA RSW, and a follow-on project in 2022 33 helped to continue and scale these interventions. Importantly, the additional financing allowed refugees and hosts living in a project implementation area to access these programs under the same eligibility criteria. Refugee beneficiaries of cash transfers were mostly expected to come from the better- integrated Central African Republic refugees in the East, Adamawa, and North regions. Nigerian refugees in the Far North would have been harder to reach due to insecurity, remoteness, and the government’s stricter encampment practices. Progress in implementation was nevertheless difficult among refugees. Six months before the end of the project in 2022, less than 40 percent of the planned beneficiaries had been enrolled. Training and awareness efforts targeting human capital improvement and improvement of households’ income generation potential had also not yet reached refugees. To improve the situation and ensure that refugee beneficiaries, too, had access to support, the refugee- facing program shifted focus away from TMO and THIMO to TMU. The TMU surpassed its beneficiary targets for refugees, reaching 8,286 refugee households, more than three times the planned figure. TMO and THIMO, however, stopped short of the target, together benefiting just under 11,000 of the planned 16,500 refugee households. Because of the shift to TMU and its lower cost per beneficiary, the overall target beneficiary number had, however, been exceeded, both for refugees and for hosts, at project completion. The more complex, income-generating activities reached just over 80 percent of the planned number of refugee beneficiaries. The project’s completion report suggests that this outcome was linked to issues with outreach and enrollment mechanisms for refugees. Among the most challenging aspects of the refugee-facing program, the project team mentioned its implementation in remote and insecure parts of the country, the need to adapt program training and communications to the different languages spoken by refugees, and a lack of refugee documentation—a concern particularly for income-generating activities and public works. Weak central government ownership of social protection 33 Launched in 2022, the Adaptive Safety Net and Economic Inclusion Project also includes the Youth Economic Inclusion and Business Plan Competition programs, which target young people in urban areas, with a particular focus on IDPs and other vulnerable groups. 29 programs and refugee assistance, and inadequate local infrastructure and implementation capacity, complicated progress. Institutional collaboration with the UNHCR, beyond receiving feedback on project implementation, was also limited, reducing possibilities for the government to learn from and take over from humanitarian experience and structures. The project’s focus on refugee communities living outside of camps enabled targeting to be independent of the household’s nationality or refugee status. This made it possible to include refugees in all project schemes without additional implementation arrangements. Findings from a qualitative assessment (Della Guardia et al. 2024) suggest that the program generally improved social cohesion within communities, as beneficiaries and nonbeneficiaries bonded through informal mechanisms of solidarity and sharing, particularly in some displaced communities. However, this was not the case among host communities in more insecure areas of Cameroon, where several beneficiaries reported that nonbeneficiaries resented them. In hindsight, the project team suspected that merely opening the project to refugees may not have been enough. A more deliberate, intensive campaign to enroll refugees might have sped up implementation and ensured that the more complex, comprehensive operations—the public works and cash transfers with accompanying measures—also reached the planned number of refugee beneficiaries. This highlights the double-edged nature of merely “opening” a transfer program to refugees. Not all communities are alike, and optimizing communications, both for outreach and to build social cohesion, is challenging, particularly when government capacity and appetite for social protection are constrained. Indeed, stronger outreach could have helped the project reach the planned number of refugee beneficiaries. While the program demonstrated feasibility, the full scale of its ambition could not be achieved in time under these circumstances. Integration of refugees and their hosts in Ethiopia’s safety net Implementation category: Government led Ethiopia is among the Sub-Saharan African countries hosting the largest number of refugees—approximately 980,000. 34 Just under half are from South Sudan, and most of the others are from neighboring Somalia, Sudan, and Eritrea. The Refugee and Returnees Service 34 UNHCR data. 30 (RRS) of the Government of Ethiopia (GoE) oversees policy and programming for refugees hosted in Ethiopia. The UNHCR provides technical and financial support. The majority of Ethiopian refugees are hosted in refugee camps, managed by the RRS. The WFP, with the financial support of donors, provides refugees with monthly rations consisting of a mix of food and cash; special support is provided to particularly vulnerable groups. Much like in Ethiopia, these rations have been cut drastically in the past few years, in response to dwindling funding (WFP 2022). In response, the World Bank and the GoE have designed Refugee and Host Integration through the Safety Net (RHISN), a WHR-funded subcomponent of the Urban Productive Safety Net and Jobs Project (UPSNJP), which extends the urban safety net to selected refugee camps and their host communities. Safety nets have a long-standing history in Ethiopia. The rural and urban safety nets— together reaching approximately 9 million individuals with cash, food transfers, and social services—make up the cornerstone of the GoE’s social protection strategy and policy. The centerpiece of the UPSNJP is the public work and livelihood development program—a three- year poverty-graduation program of the Ministry of Urban and Infrastructure that supports beneficiaries in 88 cities with cash-for-work, savings, training, and grant funding for them to pursue higher income generation in wage- or self-employment. Under RHISN, this program enrolls about 22,500 households, split according to a quota system between refugees (42%) and hosts (58%). An additional financing raised the number of households to about 34,600. The GoE’s 2019 Refugee Proclamation gave refugees the right to work in self- and wage employment (Federal Democratic Republic of Ethiopia 2019) in joint community projects such as those under the RHISN. Beneficiary refugees obtain five-year residence permits, which grant them the right to engage in wage and self-employment under the RHISN’s livelihood program. The project aims both to alleviate poverty among participants and to build social cohesion among the communities. Refugee and host beneficiaries are expected to work side by side in joint public works, cleaning and beautifying public places; planting trees; and building small urban infrastructure, such as gutters, toilets, and classrooms; and also engage in other bespoke social cohesion-building activities. The RHISN is being implemented in three phases. Phase 1 covers 10 percent of the overall target number of beneficiaries and is active in Ethiopia’s Somali Regional State since August 2022. Phase 2 covers the balance of the beneficiary pool and has been rolling out progressively in selected sites in Amhara, Benishangul-Gumuz, Gambella, and Afar Regional States—the other major refugee-hosting regions of the country. In Phase 3, five cities that 31 were already participating in the broader UPSNJP, received a further allocation to expand their operations to the refugees living within the city limits, or in nearby camps. Although closely modeled on the existing urban safety net, the integration of community mixing in public works and the presence of social cohesion activities represented substantial innovations that required thorough site-specific understanding of dynamics. Public works and livelihoods programming that bring refugees and hosts together require certain enabling conditions, including, at a minimum, physical proximity of refugee camps to host communities, good relations between hosts and refugees, and an interest within the communities to participate in such programming. Phase 1 focused on demonstrating feasibility at small scale in an area where refugees were culturally similar to hosts and had largely integrated. A socioeconomic study assessing conditions in selected camps and surrounding communities was also completed in Phase 1, though it took six months longer than anticipated due to fieldwork disruptions caused by the outbreak of widespread conflict in Ethiopia in 2020. The RHISN also has had to adapt to the specific context of refugee camps and refugee- hosting areas. While beneficiary status in the broader UPSNJP is relatively stable, the more fluid, occasionally volatile, living conditions of refugee-hosting towns, posed peculiar challenges. A bout of conflict in Pugnido—a town implementing the RHISN in the Gambella region—led to a massive dropout of the refugee beneficiary pool—over half—from the program. The program responded nimbly, reenlisting households within a few weeks. Refugees in Phase 2 areas, in particular, tended to drop out more—drawn to economic opportunities elsewhere, in artisanal gold mining further afield, in nearby cities with kindred communities, or in Addis Ababa, seeking job opportunities in a secure, diverse environment. As a result, by June 2024—a year and a half from the originally planned closure of December 2025—the RHISN had enrolled only 88 percent of its target beneficiaries, whereas the main operation had completed enrollment one year before. M&E and implementation support activities related to Phase 1 suggest four reasons why progress was harder and slower: remoteness of refugee-hosting communities, lack of information on these communities, capacity and bandwidth constraints of the implementing agencies, and other program priorities competing for the attention of program management. In Ethiopia, as elsewhere in the world, refugee-hosting areas are typically remote and economically marginalized. Refugee camps are often far from large towns—in Ethiopia, one camp is over 100 kilometers from the nearest city participating in the UPSNJP. This makes responding through an urban safety net challenging. The project originally envisaged remote 32 implementation: the UPSNJP cities would establish satellite offices in hosting communities and implement remotely. The resulting mandate issues proved intractable. After careful deliberation, authorities in Somali Regional State (Phase 1) opted to request the introduction of a new town in the UPSNJP rather than attempt remote implementation. Implementation preparations were delayed by several months as a result. The RHISN necessitated a novel, and ultimately winning, collaboration between the Ministry of Urban and Infrastructure and the RRS. The initial phases, however, stretched the institutions’ capacity at both federal and local levels. RRS, a new implementing partner in the safety net, manages the targeting of refugee communities. In Phase 1, the RRS had to determine, for the first time in its history, systems for verifying, processing, and communicating multiple iterations of ranked safety-net beneficiary lists at the local and federal levels. This necessitated thorough capacity building and bespoke institutional arrangements in a town that lacked the usual safety-net staff, training, and equipment. This meant that rollout was slower than had originally been planned. In Phase 2, a second rollout of the RHISN took place in these same towns. The process was much smoother and quicker, since the RRS had determined the level of effort involved in processing these lists and assigned the workload accordingly within its structures. This demonstrated the importance of learning-by-doing and of allowing ample lead time for such operations to roll out in novel institutional contexts based on trial, error, and adaptation. For all these reasons, the RHISN had to innovate at almost every step of the way and proceeded on a deferred rollout timeline compared with the rest of the urban safety net. Due to its smaller scale, the RHISN struggled to obtain the level of prioritization that would have allowed it to catch up to the main program. Regional awareness raising and capacity building for Phase 1 and Phase 2 proceeded on a lagged timeline compared with the main program. Phase 2 cities had hence to speed up implementation, compressing the program to two years, rather than the planned three.` Most of the project’s progress was driven by the RHISN project coordination unit – and eventually RRS with the World Bank’s technical support. Although it provided welcome additional funding for the safety net and for refugee care and maintenance, refugee integration in the safety net was not per se a government priority and did not have an obvious ministerial champion: this made it difficult to maintain the momentum of implementation. Because all the challenges were ultimately overcome, the GoE did invest further in the RHISN Phase 3, whose implementation is proceeding faster than Phases 1 and 2. Phase 2 is the subject of one of the most innovative impact evaluations currently in the 33 field, for the social protection of refugee communities, which is set to be the first large-scale RCT of a government-implemented safety-net intervention in the world. The Bangladesh portfolio for the displaced Rohingya population Implementation category: Parallel/Stand-alone, then shadow alignment Fleeing systematic violence in Myanmar that began in October 2016, over 1.1 million people from the Rohingya ethnic group have sought refuge in neighboring Bangladesh (Government of Bangladesh and UNHCR 2025). In the immediate aftermath, local NGOs, the Government of Bangladesh (GoB), Bangladeshi citizens, as well as humanitarian aid supported the settlement of the displaced, who mostly live in Cox’s Bazar, one of the least-developed districts of Bangladesh. At the invitation of the GoB, the World Bank’s IDA RSW provided support to the response, investing almost US$492 million in two phases across health; education; social protection; and basic infrastructure like water supply, sanitation, and transport. In the first phase, between 2019 and 2020, three RSW-funded operations 35 worth about US$200 million were delivered on an accelerated timeline. These operations were focused on the immediate needs of the displaced Rohingya population (DRP 36). In the second phase, starting in 2020 and worth about US$292 million, the Bank fielded a regional development approach with three new RSW-funded operations, 37 covering the whole district of Cox’s Bazar and targeting both the DRP and host communities. Mirroring the shift in focus from humanitarian response to development, the second-phase interventions emphasized establishing mechanisms for the delivery of basic services to both host communities and DRP under the same operations, and implementing them through existing government structures, where possible. 35 Namely, the Additional Financing for Health Sector Support Project (P167672), the Emergency Multi-Sector Rohingya Crisis Response Project (P167762), and the Additional Financing for Reaching Out of School Children II (P167870). 36 The GoB is not a party to the 1951 Refugee Convention. Rohingyas fleeing Myanmar to Bangladesh are therefore not refugees, in the formal, legal sense. 37 Namely, the Health and Gender Support Project for Cox’s Bazar district (P171648), the Emergency Multi- Sector Rohingya Crisis Response Project Additional Financing (P171779), and Safety Net Systems for the Poorest (AF) (IDA18 RSW) (P171086). 34 Two RSW-funded projects provided cash-for-work for the DRP. The first was a US$35 million component of a multisectoral project (World Bank 2019), approved in 2019 and concluded in July 2022. The second was US$100 million additional financing to an existing operation 38 (World Bank 2020) that closed in December 2024. COVID-19 restrictions meant that the number of physical community works under the first project had to be limited. However, by including community services, the project managed to deliver stipends to 175,000 out of an estimated 196,000 DRP households in Bangladesh at the time—75 percent more than the planned target of 100,000. The second project included a US$70 million scale-up of the Employment Generation Program for the Poorest (EGPP)—a GoB safety-net program supporting poor households in the low season for agricultural employment. The intervention, termed EGPP+, was created as a shock-responsive scale-up of the EGPP in Cox’s Bazar to help the resident Bangladeshis cope with the socioeconomic impacts of the DRP’s continued displacement. Additionally, US$30 million continued the support for the DRP, originally launched under the aforementioned component of the multisectoral project. 39 The cash-for-work activities were similar across both projects, for DRP and hosts. Men able to supply heavy manual labor received payment for participation in community works, involving small public infrastructure and environmental maintenance interventions. Meanwhile, women and persons with disabilities were remunerated for delivering community services of awareness raising around climate-resilient and family best practices, social mobilization, care, and cleaning. Though coordinated by the same ministry, the delivery systems for each population remained distinctly parallel, even in the second project. The WFP implemented the DRP facing program, based on contractual arrangements with the Ministry of Disaster Management and Relief (MoDMR), while the Department for Disaster Management, also under the MoDMR, implemented the EGPP+. Wage rates and payment modalities also differed. Hosts received payments via fully functional mobile financial services (mobile money). DRP received cash-equivalent payments on electronic vouchers managed by the WFP, which could be redeemed only in certain shops. Though it does not aim for DRPs’ integration into the social protection system, the project’s inclusion of host support alongside a DRP facing intervention achieves two intermediate 38 Namely, the Safety Net Systems for the Poorest (P171086). 39 Namely, the Emergency Multi-Sector Rohingya Crisis Response Project (P167762). 35 steps: making the safety net itself more shock responsive and linking social assistance for host populations to the presence of DRP. By tying social protection support to a shock, it enhanced the EGPP’s capacity to respond to shocks overall, moving away from the seasonal support model it had employed to date. This laid the groundwork for the shock-responsive window, EGPP+, to be mobilized in 2022 in response to floods in other districts in the country. A strengthened social protection system in Cox’s Bazar, led by local government, should be better placed to pick up from where humanitarian response leaves off, both in managing the risks and downsides and in maximizing the benefits of the DRP’s presence. This transition cannot happen spontaneously and will require appropriate grant funding from donors, smart communications with local hosting communities, and the building of government agencies’ capacity to effectively reach the Rohingya, in mutually intelligible and culturally compatible ways. While the government maintains that repatriation is the only sustainable solution, many years later, the focus should shift from preventing the worst in a humanitarian emergency to making the most, equitably, of the development challenges posed by a protracted displacement crisis. Stakeholder feedback highlighted issues with the time duration and transactional costs of procedures and approvals, impeding the Bank’s capacity to quickly transition from emergency response to development programming. Some of the challenges mentioned were: (1) delays in contracting and payment arrangements with UN implementing agencies due to lengthy GoB clearance and/or contract approval processes, and (2) emphasis on the pursuit of certain corporate targets that hindered the speed of delivery of the operation’s urgently needed assistance. In later iterations of support to the DRP, delays were reduced as the GoB, through World Bank persuasion, allowed the multiyear contracting of UN agencies, lessening the procedural burden. The second issue, that of corporate targets, manifested itself in discussions around women’s participation in physical public works. A majority of DRP do not generally encourage women to work outside the household, especially together with men (Guglielmi et al. 2022). While advancing gender norm change is an important long-term development goal, the need to deliver income support to vulnerable DRPs, especially women, required working within existing social norms as a constraint. In the end, the team devised an approach to community works that the DRP found socially acceptable for women, and the income support reached its beneficiaries. In parallel to the safety-net interventions, the health and education projects also managed to combine government stewardship and ownership with the delivery capacity of UN 36 agencies, although not without instructive challenges. Important and resolvable challenges surfaced in the humanitarian third-party implementation model. In health programming, the third-party implementation model was complicated by bureaucratic hurdles such as elaborate internal clearance procedures in the more centralized UN agencies, and internal Bank financial controls limiting US-dollar-denominated contracts to UN agencies to protect the contracts from adverse fluctuations in currency markets. Similar bureaucratic challenges were seen in education. Procurement and safeguarding rules for UN agencies and the Bank are not necessarily aligned, potentially adding delays. For example, in education, learning center designs approved by a UN agency fell short of Bank requirements. Both projects faced delays in implementation because of these limitations and the impact of COVID-19, and the education project did not complete all planned learning centers or reach the hoped-for levels of enrollment. A more radical challenge to the third-party implementation model is financial and institutional sustainability. Implementation via contracts with the United Nations offers the value of tried and tested financial management and procurement methods. It is also expensive, with overhead costs fixed at 4 percent in the social protection projects but rising into the double digits for other RSW interventions in Bangladesh. 40 UN agencies then subcontract many activities to NGOs, which introduce their own costs. With service delivery to the camps still routed through UN agencies and NGOs, the available—and shrinking— funding pool cannot cover needs. Implementation arrangements that are more directly integrated with local systems, whether government- or NGO-led, might be able to do more for less. Most Rohingyas displaced abroad wish to return to Myanmar, if living conditions are safe and conducive, while 3 in 10 simply rule it out (UNHCR 2022a). That raises an even thornier question for the refugee-facing windows: How can the Bank, together with the broader international community, better serve the wishes of this majority who wish to return to Myanmar? How can it field support and wield its influence in such a way that helps Rohingya in their quest to return home? Facilitating sustainable solutions, including safe return, is one of the objectives of IDA-19 WHR. 40 The WFP overhead was fixed at 4 percent in the completed contracts. 37 4.3 What can be learned from these experiences? At the outset, this discussion identified four dimensions along which government-led integration of refugees in safety-net operations could be expected to differ from pure humanitarian operations: capacity, ownership, cost structures, and integration. While it is too early in the implementation process—and data from these country examples are too limited—to comment on cost structures or integration, the case studies highlight important insights regarding capacity and ownership. The capacity of implementing government institutions is a particularly significant constraint. As seen in Ethiopia and Cameroon, low capacity in remote, peripheral areas slows implementation. In Cameroon, language barriers compounded the problem. By contrast, the Bangladesh case study showed how ambitious project targets can be not only met but even exceeded when projects leverage the capacity of the humanitarian community. Closely tied to capacity constraints is government ownership and mandate. In L&MICs, before a safety-net intervention reaches out to the refugee population, refugee programming is handled more often by humanitarian agencies than by the government. In some cases, like Cameroon, the government may entirely lack a specific refugee agency. Where a refugee agency exists, its mandate may focus more on the protection of refugees and the security of the state, than on development and integration. Line ministries or agencies tasked with implementing safety nets have no prior mandate, and usually no experience delivering assistance to refugees, or their host populations. A refugee-facing safety net raises the question of mandates and requires a high degree of collaboration and a new division of labor between the line ministries and refugee or humanitarian agencies. All three case studies highlighted a resulting procedural and institutional divide between refugee programming and line ministries tasked with social protection systems. In the cases reviewed, building the coordination structures needed to bridge this gap and implement an integrated safety-net system—in sharing targeting data in Ethiopia, in leveraging UNHCR support in Cameroon, and in moving past the parallel delivery system model in Bangladesh— proved difficult. 41 Political enthusiasm for including refugees in programs targeting the most vulnerable is understandably limited in L&MICs, where policy makers may well hesitate to devote resources and government time to nonnationals, when many fellow citizens are just as needy. In peripheral areas, political interest in developing social protection systems even 41 Indeed, a further round of assistance approved in 2024, across social protection, health, education, and gender-based violence response and prevention in Bangladesh, will perpetuate this parallel approach, where UN agencies implement for the DRP, while the government implements for nationals. 38 for nationals is low, and it is even lower for refugees, who risk diverting humanitarian support. The multisectoral response in Bangladesh offers another important insight. The cash transfer component of the emergency response was the first of the three interventions to be fully completed even though COVID-19 forced a redesign. Cash proved faster even though it was not fully digital, as it involved the physical distribution of e-voucher cards. This is intuitive because cash transfers need little infrastructure beyond a payment point and field presence to carry out targeting. Environmental and social standards are more easily complied with, and procurement can be minimal, in contrast to other social service delivery systems. 4. Can government-run safety-net interventions for refugees combine cost-effectiveness and impact? There are good reasons to believe that cash transfers and cash-plus approaches hold promise for improving socioeconomic outcomes among the most economically vulnerable refugees and their host communities in L&MICs. Where short-term evidence on the causal impacts of government-run IDA-funded programming is available, it shows that programs have significant, positive impacts on economic well-being, but refugees and host communities are affected to a different degree. However, the experience reviewed so far suggests that it is challenging to implement programming via host government systems. Even when the delivery system itself has been operational for a while, as in Cameroon and Ethiopia, implementation can be hindered by the capacity and ownership of the agencies responsible. The speed of rollout can in turn be affected. Humanitarian agencies, on the other hand, can field high-quality interventions to support refugees, and have a strong institutional incentive to do so quickly and effectively. However, their interventions tend to focus less on host communities and carry higher costs. If this refugee-bias in humanitarian funding persists – as it is likely to in the context of shrinking global aid budgets – implementing cash transfers through humanitarian agencies may come at the cost of lower refugee integration with host communities and lower cost-effectiveness. Further, since humanitarian funding is liable to fluctuate over time and, more consequentially, to taper off as refugee situations become protracted, humanitarian cash transfers may lack the sustainability that government delivery systems can offer to poor refugees with no near-term prospects of returning home. 39 In other words, government-run schemes offer a lower-cost, longer-term solution with greater potential for integration, while humanitarian-run schemes offer stronger technical capacity, are more closely invested in refugees’ welfare, and can be fielded quickly. The two poles of the continuum identified by Seyfert et al. (2019) between humanitarian- and national system-led schemes appear to exhibit complementary strengths (as schematized in figure 2). The higher capacity and ownership of humanitarian-led programs enable faster implementation, making them ideally suited to immediate or short-term emergency response and the rapid establishment of delivery systems. At the other end, the (likely) advantages of government-led programming make them stronger candidates for durable solutions. Toward the middle of the continuum, L&MIC governments and humanitarian agencies seek to join forces to better deliver safety-net support to refugees, through arrangements such as shadow alignment and piggybacking. These complementary strengths of delivery modalities suggest an important operational question: can the respective strengths of the delivery modalities be combined in practice? Figure 2. Emerging strengths of safety net delivery modalities for refugees and hosts The combination would have to start from examining the status quo. Does a safety net exist? Is there a standing humanitarian program for refugees? In the common case where both exist, the question becomes: how can the confluence of humanitarian response into government systems best be engineered? One option is via piggy-backing. Humanitarian agencies would help government agencies responsible for safety-net implementation to build their capacity to support refugee and host populations. Humanitarian agencies would provide analytical support in geographic and household targeting, while rebuilding their own 40 targeting, payment, and case management systems to align with government systems. In turn, the government would demonstrate stronger ownership of refugee outcomes in policy and practice, for example by legislating and enforcing measures to ensure that refugees have access to formal work authorizations and – where possible – social services. This might be most suitable where there is already a humanitarian system serving a large population of refugees in protracted displacement, and where there is higher likelihood of government abiding by international protection standards for refugees and their data. The other approach is shadow-alignment: humanitarian agencies establish a delivery chain that mimics the government’s delivery chain. In this case, humanitarian actors set up analytics, targeting, databases, payment, and case management systems to meet refugees’ specific needs in line with international protection standards. In contrast to standalone interventions, these would be fully integrated and aligned, from the outset, with government safety-net systems. At a minimum, the systems would have to be designed with the possibility to integrate at a later stage, and any additional support for refugees that the safety-net system does not grant to nationals would have to be designed, communicated, and implemented as a temporary, targeted top-up. This appears more suitable during a crisis response, when government safety net systems may be unprepared and unfunded, or where there is no political mandate to provide immediate support to an inflow of new refugees. 42 Impact evaluations of the many refugee-facing safety net operations that are still under implementation should bring key lessons to bear on the question of how to provide social protection to refugees. The literature surveyed is set to expand substantially. In the meantime, the discussion of WHR programming and the literature examined in this paper already warrants additional practical considerations and recommendations for future operations. For instance: When extending government-run safety nets to refugee-hosting areas, it is critical to budget extra time and resources for capacity building, especially in remote areas. Reaching refugees and their hosts is harder than reaching nationals. It is not likely to be a straightforward rollout and they face different conditions. Government capacity and ownership are also likely to be a stricter constraint than in “vanilla” safety nets, making it essential to give governments the time and the technical assistance needed to forge official 42 What happened in Türkiye seems to be an instance of this, where safety net assistance to Syrian refugees was started with intense technical assistance by the WFP and is now administered predominantly by Turkish actors. 41 partnerships and identify the institutional champions that can ensure that implementation retains momentum. A pilot demonstrating feasibility at a small scale and in an easy environment will go a long way in enabling wider interventions and change. Governments could leverage humanitarian agencies’ support to speed up implementation in the early stages, while planning for long-term integration with national systems. Government agencies implementing safety nets may be interested in outsourcing some of the mechanics of dealing with refugee populations. Data sharing, capacity building, and collaborations with humanitarian actors in areas such as communications, targeting, grievance redress, and tailored support for refugees and their hosts can help accelerate progress, and are areas in which humanitarian actors often have a comparative advantage. On the other hand, wherever possible, humanitarian actors should consider integration with the national social safety net as their long-term goal, and build their interventions accordingly in partnership with the host government. Implementing safety net operations to support the most economically vulnerable among refugees and their host communities represents the most systematic attempt so far to honor the GCR’s commitments to channel humanitarian aid increasingly through host government channels in L&MICs. Doing so effectively requires a new level of cooperation among humanitarian support, development aid, and government policy. More robust evidence should be gathered and analyzed on important lingering questions: What is the efficacy and the cost-effectiveness of these interventions in improving refugees’ economic prospects? What are the impacts on the integration of refugees within local communities? Does host involvement improve the impacts? Which processes and models work? Which do not work? Why? This study has presented a glimpse of the available operational experience. This paper’s limited scope points to other important avenues for operational research. Future work should consider factors that influence the suitability of each of Seyfert et al. (2019)’s models in a given context more systematically and broadly, beyond the stage of the displacement situation, delivery-system capacity, government ownership, and the need for speed. The maturity of the national safety net, the political economy of refugee-hosting, the legal and policy frameworks around refugees’ economic and social rights, and the presence and role of humanitarian actors are all important dimensions for assessing country readiness. As relevant impact and process evaluations worldwide finally start yielding results, the insights can be published and discussed widely with refugee-hosting governments and the development community making these important investments. The next investment rounds 42 are likely to make further, more systematic and farther-reaching headway into the integration of refugees within national social protection systems. 43 References Aker, Jenny C. 2015. “Comparing Cash and Voucher Transfers in a Humanitarian Context: Evidence from the Democratic Republic of Congo.” The World Bank Economic Review 31 (1): lhv055. https://doi.org/10.1093/wber/lhv055. Altındağ, Onur, and Stephen D. O’Connell. 2023. “The Short-Lived Effects of Unconditional Cash Transfers to Refugees.” Journal of Development Economics 160 (January). https://doi.org/10.1016/j.jdeveco.2022.102942. Araujo, Maria Caridad, Mariano Bosch, and Norbert Schady. 2019. “Can Cash Transfers Help Households Escape an Inter-Generational Poverty Trap?” In The Economics of Poverty Traps, 357–82. Chicago: The University of Chicago Press. Baird, Sarah, Craig McIntosh, and Berk Özler. 2019. “When the Money Runs Out: Do Cash Transfers Have Sustained Effects on Human Capital Accumulation?” Journal of Development Economics 140 (September): 169–85. https://doi.org/10.1016/j.jdeveco.2019.04.004. Bastagli, Francesca, Jessica Hagen-Zanker, Luke Harman, Valentina Barca, Georgina Sturge, Tanja Schmidt, and Luca Pellerano. 2016. “Cash Transfers: What Does the Evidence Say? A Rigorous Review of Programme Impact and of the Role of Design and Implementation Features.” London, UK: Overseas Development Institute. https://cdn.odi.org/media/documents/11316.pdf. Bold, Tessa, Mwangi Kimenyi, Germano Mwabu, Alice Ng’ang’a, and Justin Sandefur. 2018. “Experimental Evidence on Scaling up Education Reforms in Kenya.” Journal of Public Economics 168 (December): 1–20. https://doi.org/10.1016/j.jpubeco.2018.08.007. Bossuroy, Thomas, Markus Goldstein, Bassirou Karimou, Dean Karlan, Harounan Kazianga, William Parienté, Patrick Premand, et al. 2022. “Tackling Psychosocial and Capital Constraints to Alleviate Poverty.” Nature 605 (7909): 291–97. https://doi.org/10.1038/s41586-022-04647-8. Brück, Tilman, Francisca Castro, Sarah Fenzl, Wolfgang Stojetz, and Philip Verwimp. Forthcoming. The Impacts of the Lisungi Project on Refugees and Host Populations in Likouala, Republic of Congo: Midline Analysis Report. Berlin, Germany: ISDC – International Security and Development Center. Caria, Stefano, Grant Gordon, Maximilian Kasy, Simon Quinn, Soha Shami, and Alexander Teytelboym. 2021. “An Adaptive Targeted Field Experiment: Job Search Assistance for Refugees in Jordan.” CEGA Working Paper Series 172 (August). https://doi.org/10.26085/c3c012. 44 De Hoop, Jacobus, Mitchell Morey, and David Seidenfeld. 2019. “No Lost Generation: Supporting the School Participation of Displaced Syrian Children in Lebanon.” The Journal of Development Studies 55 (sup1): 107–27. https://doi.org/10.1080/00220388.2019.1687875. Della Guardia, Anne, Milli Lake, and Mira Saidi. 2024. Safety Nets in Contexts of Violence, Fragility and Forced Displacement: The Case of Burkina Faso and Cameroon. Social Protection & Jobs Discussion Paper No. 2406. Washington, DC: World Bank. http://hdl.handle.net/10986/42067. Development Initiatives. 2024. “Global Humanitarian Assistance Report - Falling Short? Humanitarian Funding and Reform.” Development Initiatives. https://devinit.org/resources/falling-short- humanitarian-funding-reform/. Devictor, Xavier, and Quy-Toan Do. 2017. “How Many Years Have Refugees Been in Exile?” Population and Development Review 43 (2): 355–69. https://doi.org/10.1111/padr.12061. Doocy, Shannon, and Hannah Tappis. 2017. “Cash‐Based Approaches in Humanitarian Emergencies: A Systematic Review.” Campbell Systematic Reviews 13 (1). https://doi.org/10.4073/csr.2017.17. Egger, Dennis, Johannes Haushofer, Edward Miguel, Paul Niehaus, and Michael Walker. 2022. “General Equilibrium Effects of Cash Transfers: Experimental Evidence from Kenya.” Econometrica 90 (6): 2603–43. https://doi.org/10.3982/ecta17945. Elice, Paola. 2021. “Impact Evaluations in Forced Displacement Contexts: A Guide for Practitioners.” Reference Paper for the 70th Anniversary of the 1951 Refugee Convention, October. https://www.unhcr.org/people-forced-to-flee-book/wp- content/uploads/sites/137/2021/11/Paola-Elice_Impact-Evaluations-in-Forced-Displacement- Contexts-A-Guide-for-Practitioners.pdf. European Court of Auditors. 2024. “Special Report 06/2024: The Facility for Refugees in Turkey.” European Court of Auditors. European Union. https://www.eca.europa.eu/en/publications?ref=SR-2024-06. European Union, DG NEAR. 2021. “Strategic Mid-Term Evaluation of the Facility for Refugees in Turkey. Final Report Volume I: Main Report.” European Union. https://neighbourhood- enlargement.ec.europa.eu/document/download/e11761f6-8b9e-4465-9e47- f49fced53a51_en?filename=Vol%20I%20-%20Main%20Report.pdf. Federal Democratic Republic of Ethiopia. 2019. Refugee Proclamation 1110/2019. Federal Negarit Gazette. 45 Fein, Frieda, and Jamon Van Den Hoek. 2022. “Do Refugee Camps Offer a Refuge from Conflict? A Spatially Explicit Analysis of Conflict Incidence at 1,543 Refugee Camps across Africa (1997–2020).” Frontiers in Human Dynamics 4 (March). https://doi.org/10.3389/fhumd.2022.857250. Ginn, Thomas, Reva Resstack, Helen Dempster, Emily Arnold-Fernández, Sarah Miller, Martha Guerrero Ble, and Bahati Kanyamanza. 2022. “2022 Global Refugee Work Rights Report.” Washington, DC: Center for Global Development. https://www.cgdev.org/sites/default/files/2022-global-refugee- work-rights-report_0.pdf. Goltz, Jan von der, Kirsten Schuettler, Julie Bousquet, and Tewodros Aragie Kebede. 2024. The Labor Market Impact of Forced Displacement. World Bank. https://openknowledge.worldbank.org/handle/10986/40701. Government of Bangladesh, and United Nations High Commissioner for Refugees (UNHCR). 2025. “Joint Government of Bangladesh - UNHCR Population Dashboard as of May 2025.” UNHCR Operational Data Portal (ODP). https://data.unhcr.org/en/documents/details/116843. Hagen-Zanker, Jessica, Marcela Rubio, Christina Lowe, and Caterina Mazzilli. 2022. “Basic Needs and Wellbeing in Displacement Settings the Role of Humanitarian Assistance and Social Protection.” London, UK: ODI. https://media.odi.org/documents/basic_needs_wp_-2022-06-29.pdf. Harvey, Paul, and Habiba Mohamed. 2022. “The Politics of Donor and Government Approaches to Social Protection and Humanitarian Policies for Assistance during Crises.” BASIC Research Working Papers 10 (March). https://doi.org/10.19088/basic.2022.010. Heisey, Janet, Inés Arévalo Sánchez, and Alexi Bernagros. 2022. “Working for Inclusion: Economic Inclusion in Contexts of Forced Displacement.” Partnership for Economic Inclusion: In Practice Notes 4. https://openknowledge.worldbank.org/server/api/core/bitstreams/f9d7b36b-d436-5c62-9f23- cdd1e0827726/content. Hidrobo, Melissa, John Hoddinott, Amber Peterman, Amy Margolies, and Vanessa Moreira. 2014. “Cash, Food, or Vouchers? Evidence from a Randomized Experiment in Northern Ecuador.” Journal of Development Economics 107 (March): 144–56. https://doi.org/10.1016/j.jdeveco.2013.11.009. Hidrobo, Melissa, Amber Peterman, and Lori Heise. 2016. “The Effect of Cash, Vouchers, and Food Transfers on Intimate Partner Violence: Evidence from a Randomized Experiment in Northern Ecuador.” American Economic Journal: Applied Economics 8 (3): 284–303. https://doi.org/10.1257/app.20150048. 46 Hızıroğlu Aygün, Aysun, Murat Güray Kırdar, Murat Koyuncu, and Quentin Stoeffler. 2024. “Keeping Refugee Children in School and out of Work: Evidence from the World’s Largest Humanitarian Cash Transfer Program.” Journal of Development Economics 168 (May): 103–266. https://doi.org/10.1016/j.jdeveco.2024.103266. Hussam, Reshmaan, Erin M. Kelley, Gregory Lane, and Fatima Zahra. 2022. “The Psychosocial Value of Employment: Evidence from a Refugee Camp.” American Economic Review 112 (11): 3694–3724. https://doi.org/10.1257/aer.20211616. Ibáñez, Ana María, Andrés Moya, María Adelaida Ortega, Sandra V. Rozo, and Maria José Urbina. 2025. "Life out of the shadows: the impacts of regularization programs on the lives of forced migrants." Journal of the European Economic Association 23, no. 3: 941-982. Imbens, Guido, and Thomas Lemieux. 2008. “Special Issue Editors’ Introduction: The Regression Discontinuity Design—Theory and Applications.” Journal of Econometrics 142 (2): 611–14. https://doi.org/10.1016/j.jeconom.2007.05.008. Innovations for Poverty Action. 2023. “The Impact of a Graduation Program on Livelihoods in Refugee and Host Communities in Uganda: Study Summary.” Poverty-Action.org. March 2023. https://poverty- action.org/node/36101/pdf. International Federation of Red Cross and Red Crescent Societies (IFRC). 2023. “IFRC Concludes Implementation of the ESSN Programme in Türkiye | IFRC.” Ifrc.org. December 6, 2023. https://www.ifrc.org/press-release/ifrc-concludes-implementation-essn-programme-turkiye. Kelley, Ninette. 2022. People Forced to Flee: History, Change and Challenge. Oxford University Press. Klugman, Jeni. 2022. “The Gender Dimensions of Forced Displacement: A Synthesis of New Research.” Washington, DC: World Bank. https://documents.worldbank.org/en/publication/documents- reports/documentdetail/895601643214591612/the-gender-dimensions-of-forced-displacement- a-synthesis-of-new-research. Kondylis, Florence, and John Loeser. 2021. “Intervention Size and Persistence.” Policy Research Working Papers 9769 (September). https://doi.org/10.1596/1813-9450-9767. Kool, Tamara A., and Zina Nimeh. 2021. “Refugees and Social Protection.” In Handbook on Social Protection Systems, 410–22. Cheltenham, UK: Edward Elgar Publishing. https://doi.org/10.4337/9781839109119.00054. 47 Lehmann, Christian, and Daniel Masterson. 2014. “Emergency Economies: The Impact of Cash Assistance in Lebanon.” https://www.rescue.org/sites/default/files/document/631/emergencyeconomiesevaluationrepor t-lebanon2014.pdf. Levine, Simon, Antoine Socpa, Nathalie Both, Heiner Salomon, and Felicien Fomekong. 2022. “Integrating Assistance to the Displaced into a Social Protection System in Cameroon an Ideal, but in Whose Interests? CASS-RT.” London, U.K.: Overseas Development Institute. https://odi.org/documents/8143/Cameroon_report-published-2022-06-09.pdf. Lind, Jeremy, Rachel Sabates-Wheeler, and Carolina Szyp. 2022. “Cash and Livelihoods in Contexts of Conflict and Fragility: Implications for Social Assistance Programming.” BASIC Research Working Papers 8 (February). https://doi.org/10.19088/BASIC.2022.008. Lind, Jeremy, Rachel Sabates‐Wheeler, and Carolina Szyp. 2023. “Cash-plus Programming in Protracted Crises: A Review of Programmes in Contexts of Overlapping Conflict, Forced Displacement and Climate-Related Shocks.” BASIC Research Working Papers 19 (June). https://doi.org/10.19088/basic.2023.001. Lindert, Kathy, Tina George Karippacheril, Inés Rodríguez Caillava, and Kenichi Nishikawa Chávez. 2020. Sourcebook on the Foundations of Social Protection Delivery Systems. Washington, DC: World Bank. https://doi.org/10.1596/978-1-4648-1577-5. Loewe, Markus, and Esther Schüring. 2021. “Introduction to the Handbook on Social Protection Systems.” In Handbook on Social Protection Systems, edited by Markus Loewe and Esther Schüring, 1–35. Edward Elgar Publishing. https://www.elgaronline.com/edcollchap- oa/edcoll/9781839109102/9781839109102.00011.xml. Lowe, Christina, and Rebecca Holmes. 2022. “Linking Humanitarian Assistance and Social Protection in Response to Forced Displacement an Analytical Framework.” https://media.odi.org/documents/Final_analytical_framework_Nov_2022_.1.pdf. Lowe, Christina, Rebecca Holmes, Caterina Mazzilli, and Marcela Rubio. 2022. “Adjusting Social Protection Delivery to Support Displaced Populations.” London, UK: ODI. https://media.odi.org/documents/Adjusting_social_protection_delivery_to_support_displaced_p opulations_2022.pdf. 48 Moussa, Wael, Nisreen Salti, Alexandra Irani, Rima Al Mokdad, Zeina Jamaluddine, Jad Chaaban, and Hala Ghattas. 2022. “The Impact of Cash Transfers on Syrian Refugee Children in Lebanon.” World Development 150 (February). https://doi.org/10.1016/j.worlddev.2021.105711. O’Brien, Clare, Zoë Scott, Gabrielle Smith, Valentina Barca, Andrew Kardan, Rebecca Holmes, Carol Watson, and Jenny Congrave. 2018. “Shock-Responsive Social Protection Systems Research: Synthesis Report.” Oxford, UK: Oxford Policy Management. https://www.opml.co.uk/files/Publications/a0408-shock-responsive-social-protection- systems/srsp-synthesis-report.pdf?noredirect=1. Özler, Berk, Çiğdem Çelik, Scott Cunningham, P. Facundo Cuevas, and Luca Parisotto. 2021. “Children on the Move: Progressive Redistribution of Humanitarian Cash Transfers among Refugees.” Journal of Development Economics 153 (November). https://doi.org/10.1016/j.jdeveco.2021.102733. Pape, Utz Johann, Theresa Parrish Beltramo, Florence Nana Pokuaah Nimoh, Laura Abril Rios Rivera, Ibrahima Sarr, and Jedediah Rooney Fix. 2021. “Understanding the Socio-Economic Conditions of Urban Refugees in Kenya : Volume c : Urban Refugees - Results from the 2020-2021 Urban Socioeconomic Survey.” Vol. C. Washington, DC, USA: World Bank. https://documents.worldbank.org/en/publication/documents- reports/documentdetail/184981636366231985/volume-c-urban-refugees-results-from-the-2020- 2021-urban-socioeconomic-survey. Pape, Utz, Benjamin Petrini, and Syedah Aroob Iqbal. 2018. “Informing Durable Solutions by Micro-Data: A Skills Survey for Refugees in Ethiopia.” https://documents1.worldbank.org/curated/en/996221531249711200/pdf/128185-WP-PUBLIC- P162987-SkillsReport.pdf. Puri, Jyotsna, Anastasia Aladysheva, Vegard Iversen, Yashodhan Ghorpade, and Tilman Brück. 2017. “Can Rigorous Impact Evaluations Improve Humanitarian Assistance?” Journal of Development Effectiveness 9 (4): 519–42. https://doi.org/10.1080/19439342.2017.1388267. Revel, Bastien. 2020. “Turkey’s Refugee Resilience: Expanding and Improving Solutions for the Economic Inclusion of Syrians in Turkey.” Atlantic Council in Turkey, UNDP. https://www.atlanticcouncil.org/wp-content/uploads/2020/07/Turkey%E2%80%99s-Refugee- Resilience-Expanding-and-Improving-Solutions-for-the-Economic-Inclusion-of-Syrians-in-Turkey- Report.pdf. 49 Roelen, Keetie, Daniel Longhurst, and Rachel Sabates-Wheeler. 2018. “The Role of Cash Transfers in Social Protection, Humanitarian Response and Shock-Responsive Social Protection.” IDS Working Paper 157 (October). https://www.ids.ac.uk/publications/the-role-of-cash-transfers-in-social- protection-humanitarian-response-and-shock-responsive-social-protection/. Sabates-Wheeler, Rachel, Jeremy Lind, Paul Harvey, and Rachel Slater. 2022. “Strengthening Responses at the Nexus of Social Protection, Humanitarian Aid and Climate Shocks in Protracted Crises: BASIC Research Framing Paper.” BASIC Research Working Papers 1 (May). https://doi.org/10.19088/BASIC.2022.001. Sabates-Wheeler, Rachel, and Carolina Szyp. 2022. “Key Considerations for Targeting Social Assistance in Situations of Protracted Crises.” BASIC Research Working Papers 12 (March). https://doi.org/10.19088/basic.2022.012. Salti, Nisreen, Jad Chaaban, Wael Moussa, Alexandra Irani, Rima Al Mokdad, Zeina Jamaluddine, and Hala Ghattas. 2022. “The Impact of Cash Transfers on Syrian Refugees in Lebanon: Evidence from a Multidimensional Regression Discontinuity Design.” Journal of Development Economics 155 (March). https://doi.org/10.1016/j.jdeveco.2021.102803. Schuettler, Kirsten, and Laura Caron. 2020. “Jobs Interventions for Refugees and Internally Displaced Persons.” Washington, DC: World Bank. https://documents1.worldbank.org/curated/en/467251592598151781/pdf/Jobs-Interventions- for-Refugees-and-Internally-Displaced-Persons.pdf. Seyfert, Karin, Valentina Barca, Ugo Gentilini, Manjula Luthria, and Shereen Abbady. 2019. “Unbundled: A Framework for Connecting Safety Nets and Humanitarian Assistance in Refugee Settings.” Social Protection and Jobs Discussion Papers, no. 1935 (September). https://documents1.worldbank.org/curated/en/970701569569181651/pdf/Unbundled-A- Framework-for-Connecting-Safety-Nets-and-Humanitarian-Assistance-in-Refugee-Settings.pdf. Türk Kızılay, International Federation of the Red Cross, and CashHub. 2021. “Emergency Social Safety Net (ESSN): Social Assistance for Refugees in Türkiye Background: Refugees in Türkiye.” https://platform.kizilaykart.org/en/Doc/rapor/CashHub_ESSN_SocialAssistanceforRefugeesinTu% CC%88rkiye_v4-2-1.pdf. 50 United Nations. 2018. “Global Compact on Refugees.” New York, N.Y.: United Nations. https://globalcompactrefugees.org/sites/default/files/2019- 12/Global%20compact%20on%20refugees%20EN.pdf. United Nations High Commissioner for Refugees (UNHCR). 2007. Handbook for Emergencies. Office of the United Nations High Commissioner for Refugees. Geneva, Switzerland: United Nations High Commissioner for Refugees. https://digitallibrary.un.org/record/597573?ln=en. ———. 2016. “Nationality of Arrivals to Greece, Italy and Spain January 2015 - March 2016.” UNHCR Operational Data Portal (ODP). May 1, 2016. https://data2.unhcr.org/en/documents/details/47393. ———. 2022a. “Intention Survey of Rohingya Refugees from Myanmar: A Regional Perspective - 2021.” Regional Bureau for Asia and the Pacific. https://microdata.unhcr.org/index.php/catalog/672. ———. 2022b. “Bangladesh - Operational Update - June 2022.” July 7, 2022. https://data.unhcr.org/en/documents/details/94097. ———. 2024. “UNHCR Bangladesh - Operational Update September 2024.” UNHCR Operational Data Portal (ODP). December 5, 2024. https://data.unhcr.org/en/documents/details/112916. ———. 2025. “Global Trends: Forced Displacement in 2024.” Copenhagen, Denmark: United Nations High Commissioner for Refugees. Valli, Elsa, Amber Peterman, and Melissa Hidrobo. 2019. “Economic Transfers and Social Cohesion in a Refugee-Hosting Setting.” The Journal of Development Studies 55 (sup1): 128–46. https://doi.org/10.1080/00220388.2019.1687879. Voluntas. 2021. “BaNVA (Basic Needs and Vulnerability Assessment) for Refugees Hosted by Sudan - September 2021.” UNHCR Operational Data Portal (ODP), https://data2.unhcr.org/en/documents/details/88840. World Bank. 2018. “Cameroon - Social Safety Net Project: Additional Financing Project Paper.” Washington, DC, USA: World Bank Group. http://documents.worldbank.org/curated/en/433291525399293923/Cameroon-Social-Safety- Net-Project-additional-financing. ———. 2019. “Bangladesh - Emergency Multi-Sector Rohingya Crisis Response Project Appraisal Document.” Washington, DC, USA: World Bank Group. https://documents.worldbank.org/en/publication/documents- 51 reports/documentdetail/655111552269808019/bangladesh-emergency-multi-sector-rohingya- crisis-response-project. ———. 2020. “Bangladesh - Safety Net Systems for the Poorest : Additional Financing Project Paper.” Washington, DC, USA: World Bank Group. https://documents.worldbank.org/en/publication/documents- reports/documentdetail/701121585965827747/bangladesh-safety-net-systems-for-the-poorest- additional-financing. ———. 2023. “World Development Report 2023: Migrants, Refugees, and Societies.” World Bank. Washington, DC, USA: World Bank Group. https://www.worldbank.org/en/publication/wdr2023. ———. 2024a. “High Income | Data.” The World Bank | Data. World Bank Group. 2024. https://data.worldbank.org/income-level/high-income. ———. 2024b. “Low & Middle Income | Data.” The World Bank | Data. World Bank Group. 2024. https://data.worldbank.org/income-level/low-and-middle-income. ———. 2024c. “IDA-21 Policy Package: The ‘Focus Areas’ Paper.” Washington, DC, USA.: World Bank Group. https://documents1.worldbank.org/curated/en/099060424223583110/pdf/BOSIB108d2e9ef0e6 1b6761912fc83146ac.pdf. World Food Programme (WFP). 2022. “WFP, UNHCR, RRS Appeal for Funding to Continue Feeding over 750,000 Refugees in Ethiopia.” August 9, 2022. https://www.wfp.org/news/wfp-unhcr-rrs-appeal- funding-continue-feeding-over-750000-refugees-ethiopia. ———. 2025. “Cameroon - Annual Country Report 2024.” https://docs.wfp.org/api/documents/WFP- 0000167052/download/. Zhou, Yang-Yang, and Andrew Shaver. 2021. “Reexamining the Effect of Refugees on Civil Conflict: A Global Subnational Analysis.” American Political Science Review 115 (4): 1175–96. https://doi.org/10.1017/s0003055421000502. 52 Social Protection & Jobs Discussion Paper Series Titles FY 2026 No. Title 2601 Refugees and social assistance in low- and middle-income countries: A review of operational experiences To view Social Protection & Jobs Discussion Papers published prior to 2021, please visit www.worldbank.org/sp. Official Use Only ABSTRACT This paper gathers lessons from major development investments and reviews the broader evidence base on the inclusion of refugees and their host communities in national social protection programs in low- and middle-income countries (L&MICs). L&MICs host the global majority of refugees, who often live below the poverty line and in impoverished host communities with inadequate access to basic services. International institutions such as the World Bank and the European Union have invested several billions of dollars to reverse this situation through social assistance programs. Growing evidence suggests cautious optimism about the results of such interventions. Although L&MICs are rolling out safety nets to refugees and their host communities, these initiatives face unique operational challenges. Policy makers should allocate adequate time and resources to build the capacity of government mechanisms for social assistance delivery in the often-peripheral areas hosting refugees, strengthen collaboration with government and humanitarian refugee agencies, and foster host governments’ buy-in to this critical agenda. JEL CODES F22, I38, I32. KEYWORDS Refugees, social protection, social assistance.