Co-authors: Dennis Egger (University of Oxford), Alfredo Manfredini-Böhm (Tony Blair Institute), Sandra V. Rozo (World Bank), and Daksh Walia (University of Oxford)
2026 Africa Evidence Summit, Addis Ababa, Ethiopia
Belay Mohammed of the University of Oxford presented a study entitled Refugee and Host Integration through the Safety Net: Experimental Evidence from Ethiopia. Conducted jointly with Dennis Egger, Alfredo Manfredini-Böhm, Sandra V. Rozo, and Daksh Walia, the study examined whether integrating refugees into Ethiopia's national social protection system could improve the well-being of both refugees and host communities while strengthening social cohesion. Using a large-scale randomized controlled trial, the research evaluated the Refugee and Host Integration through the Safety Net (RHISN) programme, providing one of the first rigorous experimental assessments of an integrative approach to refugee assistance in Ethiopia.
Mr. Mohammed began by describing the rapidly expanding global displacement crisis and its implications for host countries. Drawing on data from the United Nations High Commissioner for Refugees (UNHCR) and the United Nations Office for the Coordination of Humanitarian Affairs (OCHA), he explained that the number of forcibly displaced people worldwide had reached approximately 122 million by 2025. Despite this unprecedented level of displacement, fewer than three percent of displaced people return to their countries of origin each year. Moreover, approximately 67 percent of refugees remain displaced for five years or longer, indicating that displacement has increasingly become a long-term rather than temporary phenomenon.
He further observed that approximately 67 percent of all refugees are hosted by low- and middle-income countries, placing considerable economic and social pressure on countries that already face significant development challenges. At the same time, humanitarian financing continues to fall short of global needs. According to recent estimates, only about 45 percent of humanitarian funding requirements are currently being met. Given these realities, he argued that traditional humanitarian assistance alone cannot provide a sustainable solution. Instead, promoting refugee self-reliance has become essential for both refugees and host governments, a conclusion supported by previous research undertaken by Rozo and Moya (2025) and Rozo (2025).
Mr. Mohammed then introduced two contrasting approaches to supporting refugee self-reliance. The first is the parallel model, under which refugee-specific assistance programmes operate separately from national public services. Under this humanitarian model, refugees receive separate social protection programmes, health care, education, and livelihood support, with relatively limited interaction between refugees and host communities.
The second approach, which formed the basis of the study, is the integrative model. Under this development-oriented model, refugees are incorporated into national systems of social protection, health care, education, and other public services alongside host populations, consistent with the objectives of the Global Compact on Refugees (2018). Rather than providing separate services, the integrative model encourages joint participation of refugees and host communities within shared institutions. Mr. Mohammed explained that the study sought to determine whether this integrated approach could successfully improve outcomes for both groups and whether refugees nevertheless require additional targeted support beyond integration into national systems.
The study therefore examined the integration of refugees into Ethiopia's national public works and livelihoods-based social safety net. According to Mr. Mohammed, the research was organized around three central questions. The first concerned household well-being, asking whether participation in the programme improved the economic and psychological well-being of refugee and host households, whether impacts differed between the two groups, and whether these differences justified differentiated policy support.
The second question examined social cohesion. Specifically, the researchers investigated whether sustained interaction between refugees and host community members through shared public works activities and common institutions increased trust, strengthened social relationships, and improved attitudes between the two groups.
The third question focused on local spillover effects. The study explored whether variation in programme coverage across neighbourhoods affected non-beneficiary households, local labour markets, enterprises, prices, and broader community cohesion.
Turning to the Ethiopian context, Mr. Mohammed explained that Ethiopia remains one of Africa's largest refugee-hosting countries, accommodating more than one million refugees, primarily originating from South Sudan, Somalia, Eritrea, and Sudan. Approximately three-quarters of these refugees reside in camps, many of whom have experienced prolonged displacement. Around 60 percent are children, highlighting the long-term developmental implications of refugee hosting.
He noted that Ethiopia has adopted an increasingly inclusive refugee policy framework. In particular, Directive No. 1019/2024 expanded refugees' legal rights to work and engage in self-employment. However, despite these progressive legal reforms, implementation has remained limited, with only about five percent of eligible refugees taking advantage of these expanded employment opportunities. This gap between policy and practice provided an important rationale for evaluating more comprehensive integration programmes.
Mr. Mohammed then introduced the Refugee and Host Integration through the Safety Net (RHISN) programme. He explained that the RHISN programme represents the first expansion of Ethiopia's established national safety net into refugee camps and neighbouring host communities. It builds upon Ethiopia's long-standing Productive Safety Net Programme, which has operated in rural areas since 2005 and currently serves approximately eight million beneficiaries, as well as the Urban Productive Safety Net Programme introduced in 2016, which supports approximately 2.4 million beneficiaries across 88 urban centres.
Under the RHISN expansion, approximately 120,000 beneficiaries, representing around 10 percent of Ethiopia's refugee population, were enrolled. The programme followed a graduation model designed to promote sustainable livelihoods. During the first year, beneficiaries participated in remunerated public works activities while receiving livelihood training. During the second year, participating households received a US$600 livelihood grant, together with coaching and continued livelihood support intended to facilitate economic self-reliance.
One of the programme's defining characteristics was the integration of refugees and poor host community members into common programme structures. Refugees participated alongside host households in public works activities, and in some locations they were assigned to mixed work teams that deliberately combined members of both groups. In addition, selected groups participated in structured social cohesion activities, including shared tea gatherings and community dinners following public works sessions.
Mr. Mohammed explained that programme eligibility differed slightly between refugees and host communities. Poor host households were selected using Proxy Means Testing (PMT), while all refugee households were considered eligible for participation. Within each household, family members selected the individual who would participate in the programme.
He emphasized that participation represented a substantial economic intervention. Beneficiaries devoted up to 25 hours of work per week, while programme transfers amounted to approximately 28 percent of average household consumption among comparison households. At the community level, the intervention was equally significant, with approximately 35 percent of households enrolled overall, although programme saturation varied across neighbourhoods from 16 percent to 66 percent.
The presenter next described the study's experimental design. Four study sites across Ethiopia participated in the evaluation. The first stage involved the random selection of beneficiary households from 180 eligible neighbourhoods, referred to as ketenas in host communities and blocks within refugee camps. Neighbourhoods were randomly assigned either high programme saturation, where approximately 75 percent of eligible households received the programme, or low programme saturation, where only 45 percent were selected. Within each neighbourhood, eligible households were then randomly assigned either to receive the RHISN programme or to remain in the control group.
The second stage involved randomly assigning programme beneficiaries to different public works groups. In Dabat and Asayita, where local tensions made social integration particularly relevant, some participants worked in mixed refugee-host groups, while others were assigned to host-only or refugee-only work groups. Certain mixed groups consisted of equal numbers of refugees and hosts, whereas others included refugees as a minority. Furthermore, approximately one-third of all public works groups were randomly selected through a public lottery to participate in additional after-work social cohesion activities, such as shared tea sessions and community dinners.
Mr. Mohammed then outlined the study's extensive data collection efforts. Researchers initially listed approximately 23,000 households, using proxy means testing for host communities and administrative refugee databases for camp residents. From these, approximately 12,000 households participated in baseline and midline household surveys, although baseline data were unavailable for Dabat because of security constraints.
In addition to household surveys, the research collected detailed information from approximately 5,000 enterprises, conducted nine market censuses, surveyed market sellers, collected prices for around 100 different goods, and observed 133 public works groups using both quantitative and qualitative methods. Group observations included structured interviews, behavioural games, and direct field observations to capture patterns of interaction between refugees and host communities. Mr. Mohammed clarified that the findings presented during the conference were based primarily on results from the midline household survey.
Before presenting the findings, he emphasized the study's strong methodological credibility. The randomized controlled trial had been pre-registered through the American Economic Association's RCT Registry (Registration No. 0012288). Baseline analysis confirmed that treatment and control groups were statistically balanced across both demographic and primary outcome variables. Moreover, the research achieved exceptionally high follow-up rates, successfully tracking more than 93 percent of participating households during the midline survey. Statistical analysis also found no evidence of systematic or non-random attrition, thereby strengthening confidence in the validity of the estimated programme impacts.
Finally, Mr. Mohammed briefly explained the empirical strategy adopted by the research team. The first set of statistical models estimated the intention-to-treat effects of assignment to the RHISN programme together with the effects of varying neighbourhood programme saturation on household outcomes. A second set of models examined the influence of inter-group contact, estimating the effects of participation in mixed refugee-host work groups and varying levels of neighbourhood exposure to members of the other group. Through these analyses, the researchers sought to identify not only the direct effects of programme participation but also the broader social consequences of integrating refugees and host communities within shared public institutions.
Mr. Mohammed concluded this part of the presentation by noting that the RHISN programme represents one of the largest and most rigorously evaluated attempts to integrate refugees into a national social protection system in Africa. He suggested that the evidence generated by the study would contribute significantly to ongoing policy debates regarding refugee self-reliance, social protection, and host-refugee integration, while informing future efforts to design inclusive development programmes that benefit both displaced populations and the communities that host them.
Continuing his presentation, Mr. Belay Mohammed discussed the implementation of the Refugee and Host Integration through the Safety Net (RHISN) programme and assessed the extent to which the intervention was delivered as originally designed. He reported that overall programme compliance was high. More than 85 percent of beneficiaries were found to be participating in their assigned public works groups during unannounced spot-check visits, indicating that most participants adhered to the programme's implementation protocols. Nevertheless, he acknowledged that implementation quality varied considerably across the different study regions. Differences in programme fidelity reflected variations in local administrative capacity, operational challenges, and beneficiary complaints encountered during implementation. These regional differences, he noted, were considered carefully in the interpretation of programme impacts.
Mr. Mohammed also described the characteristics of household members selected to participate in the programme. He explained that beneficiary households tended to nominate individuals who were relatively older, more likely to be women, and generally less attached to the labour market prior to the intervention. This pattern suggested that households strategically selected members whose participation in the programme would impose the lowest opportunity cost while maximizing household benefits.
Turning to the direct economic effects of the programme, Mr. Mohammed reported that RHISN substantially increased labour supply while generating only limited displacement of existing employment. Programme participation reduced the amount of time beneficiaries spent searching for work by approximately 18 percent, while the unemployment rate declined by roughly six percentage points. These findings indicated that the programme successfully created productive employment opportunities rather than merely replacing existing jobs.
However, he explained that some degree of labour market crowding out was observed, particularly among host-community participants. Since host households were generally more economically active before joining the programme, participation occasionally displaced existing employment activities. Nevertheless, the evidence suggested that this displacement was relatively mild and selective. Rather than withdrawing from productive employment altogether, many beneficiaries chose to leave lower-paid jobs in favour of the better remunerated public works opportunities provided through RHISN. Wage employment analyses further confirmed that participants primarily exited lower-quality employment rather than reducing their overall engagement in economic activities.
Mr. Mohammed then presented evidence regarding household income. The study found that programme participation generated significant increases in household income for both refugees and host communities. Importantly, the magnitude of these gains was broadly similar across the two groups, suggesting that integrating refugees into Ethiopia's national safety net produced comparable economic benefits regardless of beneficiaries' legal status. One notable difference, however, concerned refugee households' receipt of remittances. Refugees experienced a reduction in net remittance income of approximately US$28.50 (PPP), which the researchers interpreted as evidence that improved local economic opportunities reduced dependence on financial support from relatives and external networks.
The presenter next discussed the programme's effects on household wealth and living standards. RHISN participants experienced meaningful improvements in asset accumulation and household consumption, demonstrating that increased programme income translated into stronger household economic security. Refugees and host households exhibited remarkably similar improvements in these economic indicators, reinforcing the conclusion that integrated social protection programmes can generate comparable material benefits across both populations.
Beyond these economic gains, Mr. Mohammed emphasized that the programme also produced important improvements in broader household well-being. Although both refugees and hosts benefited, the positive impacts were generally larger among host-community households. He noted that the researchers found no evidence that improvements in well-being simply reflected better public amenities in treatment areas compared with control communities. Instead, the observed improvements were directly associated with programme participation itself.
The evaluation documented positive impacts across multiple dimensions of well-being. Participating households experienced improvements in food security, indicating greater access to sufficient and nutritious food. Health outcomes also improved, accompanied by measurable gains in mental health, suggesting that increased economic stability reduced psychological stress and anxiety among beneficiaries. Furthermore, the programme strengthened women's empowerment, while improvements in children's health and developmental outcomes suggested that the programme generated important intergenerational benefits extending beyond the immediate participants. According to Mr. Mohammed, these findings demonstrated that comprehensive social protection programmes can influence multiple aspects of household welfare simultaneously rather than merely increasing income.
The presentation then shifted to one of the study's central objectives: examining whether integrating refugees and host communities within a common social protection programme could improve social cohesion. Mr. Mohammed reported that the RHISN programme significantly improved host communities' attitudes and behaviours toward refugees, even in locations where beneficiaries were not assigned to mixed public works groups. Participation in a shared national programme appeared to reduce prejudice and strengthen acceptance of refugees among host populations. Host community members expressed more favourable attitudes toward refugees and demonstrated greater support for refugee-related public policies, suggesting that integrated service delivery can strengthen positive intergroup relations.
He further explained that deliberately mixing refugees and host community members within public works groups generated additional positive effects, particularly among host participants. Mixed work teams increased opportunities for direct interaction, encouraging greater trust, stronger interpersonal relationships, and more frequent economic exchanges between refugees and local residents. Participants also reported improved social interactions, stronger community cohesion, and greater support for refugees' political and social integration within Ethiopian society.
However, Mr. Mohammed emphasized that implementing mixed work groups proved challenging in certain locations because of existing community tensions. Consequently, the benefits of structured mixing appeared to depend heavily on local context and careful programme implementation. While integration generally improved social cohesion, strategies involving direct interaction between refugees and hosts require careful adaptation to local social and political conditions.
Another important finding concerned the programme's broader community effects. Mr. Mohammed reported that the study identified substantial spillover effects arising from increased exposure to members of the other group. Neighbourhoods characterized by greater interaction between refugees and host communities experienced improvements that extended beyond direct programme beneficiaries. These spillover effects influenced social attitudes, patterns of interaction, and broader community relationships, demonstrating that integrated social protection programmes can generate positive externalities throughout entire communities rather than benefiting only participating households.
Summarizing the preliminary findings, Mr. Mohammed highlighted two overarching conclusions. First, the programme generated large economic and well-being benefits for participating households. Although some limited crowding out of labour supply occurred among host-community beneficiaries, this effect was relatively small and largely reflected movement away from lower-paying employment opportunities. Increased household savings and asset accumulation suggested that the programme could produce lasting improvements in household resilience beyond the duration of the intervention. Moreover, improvements in children's health and development indicated that the programme may generate significant long-term intergenerational benefits.
Second, the study found substantial positive effects on social cohesion. Contrary to previous research suggesting that parallel refugee assistance programmes may unintentionally increase tensions between refugees and host communities, the RHISN programme demonstrated that integrating refugees directly into existing national social protection systems significantly strengthened social cohesion, particularly by improving host communities' attitudes toward refugees. These findings suggested that inclusive national programmes may provide a more effective framework for refugee integration than maintaining separate humanitarian assistance systems.
Mr. Mohammed also emphasized that while deliberate mixing of refugees and host communities produced additional positive outcomes, successful implementation depends on local conditions. In areas characterized by existing social tensions, careful planning and community engagement remain essential for ensuring that integrated activities strengthen rather than undermine community relationships.
Concluding his presentation, Mr. Mohammed outlined several priorities for future research. The research team intends to continue evaluating programme impacts following the completion of the graduation phase, when participating households will have received the large livelihood grant and completed the full intervention. Future analyses will examine the programme's longer-term economic impacts, broader local market spillovers, and general equilibrium effects on surrounding communities. The researchers also plan to investigate productivity and group dynamics within public works teams, together with the ways in which social relationships and behavioural changes spread through community networks over time.
In closing, Mr. Mohammed expressed sincere appreciation to Ethiopia's Ministry of Urban and Infrastructure and the Refugees and Returnees Service for their close collaboration throughout the impact evaluation. He also acknowledged the generous financial support provided by the Innovations for Poverty Action (IPA), J-PAL, the Hilton Foundation, the Prospects Partnership of the World Bank, and the World Bank Research Support Budget, noting that their contributions made the research possible. He concluded by expressing hope that the evidence generated through the RHISN evaluation would contribute to the design of more inclusive and effective refugee policies that promote both self-reliance and stronger social cohesion in refugee-hosting countries.