Introduction

In 2020, 42 per cent of the 82.4 million forcibly displaced people worldwide were children, and almost half of all refugee children were out of school (UNHCR 2020, p. 6, 2021, p. 6). Despite a growing number of education interventions targeting refugee learners over the past years, access and quality remain significant challenges. Access to education remains uneven across regions, between camp and urban settings, between different operations within the same country, and between boys and girls (Dryden-Peterson 2011, p. 32). Refugee children, especially girls, lack access to post-primary education (UNHCR 2021). Access is partly conditioned by the national legislation of host states (Dryden-Peterson 2011). While some countries, such as Uganda, allow refugees to access public education, others, like Bangladesh, only offer informal education for non-registered refugees (Dupuy et al. 2022). Furthermore, many refugee children are living in learning poverty, i.e. despite being in school, they are unable to read or to understand a simple text by the age of 10, or do not achieve basic numeracy skills (Piper et al. 2020; Saavedra and Bousquet 2020). These challenges significantly endanger the achievement of the fourth Sustainable Development Goal (SDG 4), which aims to “ensure inclusive and equitable quality education and promote lifelong learning opportunities for all” by 2030 (UN 2015, p. 19). Building up for some time already, the education crisis has only been amplified by the COVID-19 pandemic (World Bank 2022).

The past few years have seen an upward trend in humanitarian funding for education (Dupuy et al. 2019), while a marked increase in different education interventions and non-governmental organisations (NGOs) is accompanied by increasing pressure from funders to provide empirical evidence regarding the impact of their interventions. This state of affairs prompted our overarching research question, What robust quantitative evidence exists regarding interventions that improve access to education and quality learning for refugee children? Briefly, our scoping review of the existing literature, presented in more detail below, revealed that there appears to be little rigorous quantitative research testing the impact of specific interventions to improve education for refugees. Most studies appear to be observational, with a staggeringly low number of empirically robust quantitative studies. We reviewed 1,873 peer-reviewed journal articles published in the period 1990–2021, only eight of which were rigorous quantitative studies that examined the impact of various educational interventions for refugee children.

While literature reviews on refugee education have been conducted before (e.g. Hamilton and Moore 2004), these have either focused exclusively on high-income countries (McBrien 2005) or have examined only higher education (Ramsay and Baker 2019; Streitwieser et al. 2019). Other reviews have analysed interventions in emergency settings more broadly (i.e. beyond situations of forced displacement) and included both qualitative and quantitative studies (Burde et al. 2015, 2017, 2019). In our own scoping review presented here, we zoomed in on education for refugees employing a strict methodological criterion to only include studies which provide robust quantitative evidence regarding the causal impact of various programmes. Our rationale for doing so is that experimental studies can be powerful scientific tools to generate evidence to inform policies. Randomised controlled trials (RCTs) are considered to be the “gold standard” of impact analysis, providing information on the causal impact of specific interventions since participants are assigned randomly to a treatment or a control group, and these groups are followed across time and compared (Riddell and Niño-Zarazúa 2016).

Quasi-experimental or experimental research

While conducting quasi-experimental or experimental research in situations of forced displacement is challenging in terms of security-related, ethical and practical reasons, the very low number of solid empirical quantitative studies on refugee education is alarming. In the worst case, while donors increase their funding for refugee education interventions, they might risk financing, in the absence of robust impact evaluations, programmes which have little or no effect on improving refugee children’s access to education or quality learning. Robust evidence from impact evaluations is therefore critical to establishing that the effects found can confidently be attributed to the intervention and not to other factors. Such systematic evidence is crucial for donors so that they fund programmes that actually work, and which in turn can improve the lives of refugee children.

The scarce robust experimental and quasi-experimental evidence we did find to be available focuses mostly on programmes implemented in high-income countries and in school settings. This does not, however, reflect the reality on the ground, since most refugees reside in low- and middle-income countries and many of them can access education only in camps. Furthermore, we do not know enough about the long-term impact of interventions, or how the same programme might have different impacts on newly arrived refugee children compared to settled refugee children. The interventions reviewed, however, do seem to point to several promising results: language acquisition – one of the most important barriers – can be improved by language learning-enhanced physical education classes, enrolment in early childhood education programmes, and through online, game-based solutions. Attendance in child and youth learning centres (CYLCs) improved literacy and numeracy skills for both girls and boys, and cash transfer programmes increased attendance rates. However, given the low and heterogeneous number of studies, it is hard to say to what extent the findings may be applicable across different countries and contexts.

Rigorous qualitative and descriptive studies

Importantly, quasi-experimental and experimental studies also have their limitations. The validity of the results from such studies is intrinsically tied to the assumptions the research is based on and the strength of those assumptions. Furthermore, these methods are not able to unpack the mechanisms underlying the overall treatment effects (i.e. how changes in the dependent variable(s) occur). Rigorous qualitative and descriptive studies can and should therefore play an important role in filling these gaps (Gopalan et al. 2020, pp. 232–233). Qualitative methods are able to provide a deeper understanding of the context and how individuals experience and perceive various interventions. Focus groups can be appropriate qualitative tools to uncover, discuss and understand commonly held perceptions and attitudes; individual interviews open up researchers’ insights into privately held opinions or experiences (e.g. Busch et al. 2021; Hattar-Pollara 2019); and in participatory research, education communities work together to better understand and respond to problems in their day-to-day work, thereby contributing to programme quality and sustainability (e.g. Cappella et al. 2011).

In the next section of this article, we provide an overview of the barriers associated with access to education and quality learning for refugees, and potential interventions that can address them. This is followed by a description of our search method and data collection. After presenting the articles we selected for review, we offer a discussion of our findings. We conclude our article with some implications for policy and needs for future research.

Refugee education: barriers to access and quality learning

Refugees face several challenges when it comes to their access to education. These barriers vary between camp and urban settings and between girls and boys. In the case of urban settings, host states might have restrictive regulatory frameworks regarding the inclusion of refugees in national education systems (Dupuy et al. 2022). Refugee children often lack necessary identification (ID) documents and certificates, making enrolment virtually impossible. Insufficient infrastructure, crowded classrooms, and the lack of appropriate hygiene facilities can prevent children, especially girls, from going to school. Distance from schools and safety issues related to travel are also major concerns, again, especially for girls. Even if education is free, indirect costs of schooling, such as uniforms and school material, are often impossible for refugee families to finance. Since refugee children often do not speak the language of the host country, linguistic challenges remain one of the most important barriers in enrolment or placing children in age-appropriate grades. When refugee children are unable to speak the language of instruction, their academic progress can be derailed, potentially leading to dropout. There might be no education opportunities beyond the primary level, and the lack of parents’ income often means that they require their child to work instead of attending school. Even those refugee students who were able to enrol in public schools might face discrimination and violence at the school, leading to high dropout rates. Furthermore, certain belief systems, such as child marriage, or the prioritisation of boys’ educations over girls, are major reasons for dropout (Burde et al. 2015; Mendenhall et al. 2017; UNHCR 2009a, pp. 9–11).

The most typical ways of measuring quality of education focus either on inputs (e.g. student–teacher ratios, percentage of qualified teachers, the extent to which certificates are recognised) or on learning outcomes (i.e. academic performance) (Dryden-Peterson 2011, pp. 31–32). However, quality can also be measured in terms of “process” (including participation), “perceptions” or “impact” (e.g. occupational opportunities) (Millán et al. 2019). Delivering quality education for refugee children can be challenging because schools often lack the necessary infrastructure and teaching materials, or there is a high student–teacher ratio, moreover, there are often only a few female teachers. Available teachers are often not well trained to consider the specific needs of refugee children, or they do not speak the refugee children’s first language. Discrimination and language barriers not only impact access, but also the quality of education (Burde et al. 2015).

Typical interventions aimed at increasing refugees’ access to education include information campaigns, accelerated education programmes (AEPs), technology-based solutions for distance learning, cash transfer programmes, opening an afternoon shift in public schools, the provision of school meals, uniforms and school materials, safe transportation, intense language education or bilingual curricula (Burde et al. 2015; UNHCR et al. 2019). When it comes to addressing barriers to quality learning, potential interventions include involvement of teachers from the refugee community, lowering the student–teacher ratio, grouping of students according to their academic performance, provision of teaching materials in multiple languages, the establishment of community centres for out-of-school children, parent-teacher associations (PTAs), and technology-based solutions (Burde et al. 2015; UNHCR 2009b).

While there is a plethora of specific education interventions, there is considerable variation when it comes to whether these programmes were evaluated or not, and if evaluated, what method was used. For example, an interview-based evaluation of an Australian school support programme called Ucan2Footnote 1 found that the programme raised awareness amongst school staff of the specific educational needs of refugee students which led to an increase in the use of interpreters and introduced additional support through the enrolment period (Block et al. 2013, p. 1350). While qualitative evaluations are important, our review focused on quasi-experimental and experimental studies because these research designs can allow for the determination of causality and are able to isolate the key factors influencing the outcome of a particular intervention.

Literature review: methodology

The main aim of our scoping review was to identify what robust quantitative evidence exists on the effectiveness of interventions aiming to improve access to and quality of education for refugee children. Our search strategy was guided by the following two research questions:

  • What interventions can increase refugee children’s school attendance and enrolment rates?

  • What interventions can impact positively the quality of education that refugee children receive?

To identify studies which evaluate the impact of interventions on education access and quality learning for refugee children, we searched in three major electronic databases (Web of Science [WoS], JSTOR and Google Scholar) and then employed a pearl growing search strategy. In the pearl growing process we reviewed the bibliography of identified articles to supplement the number of candidates selected for the final review. We stopped running our search in September 2021. We used Boolean operators, truncation (*), and searched for the keywords and their combinations, shown in Table 1.

Table 1 Keywords included in the literature search
Full size table

We applied the following inclusion criteria:

  • English language, peer-reviewed article published between 1990 and March 2021

  • Interventions targeted at refugees, forced migrants, and internally displaced persons (IDPs) below the age of 18 years

  • Evaluations of specific interventions which addressed refugee children’s access to education and/or learning quality

  • Studies that were using a quantitative methodology, quasi-experimental or experimental research design, preferably randomised control trials

  • Interventions delivered in schools, refugee camps or community settings

It is important to note that among extant rigorous quantitative studies, only a few examine the effectiveness of access and quality of refugee education interventions, while there are a number focusing on psychosocial interventions. We excluded studies which examined interventions aiming to improve the psychosocial well-being of refugee children because several systematic literature reviews focusing on such interventions have already been conducted (e.g. Frounfelker et al. 2020; Simenec and Reid 2020; Sullivan and Simonson 2016; Tyrer and Fazel 2014).

Figure 1 shows our quorum flowchart. In total, by September 2021, we identified 1,873 articles through our abstract search method. First, we read the abstracts and categorised articles as 0 = excluded, 1 = included, 99 = to be decided. Both of us coded the articles independently. We compared our findings and discussed cases when we coded differently. Inter-coder reliability was high, and discrepancies were resolved by discussion. In the first round we screened the abstracts of the 1,873 articles and selected 39 potential articles for full review. Of the excluded studies, most did not evaluate an intervention; did not sample refugees; or did not apply a quantitative research strategy. Next, we reviewed the entire text of each of the 39 selected articles, only five of which proved to meet our strict methodological inclusion criteria.

Fig. 1
figure 1

Quorum flowchart

Full size image

Due to the low number of studies resulting from this last step, we applied two additional search strategies to our shortlist of 39 articles. First, we conceded some flexibility regarding the robustness of the study. For example, studies were accepted if they had a control and intervention group even if the participants were not randomly assigned to these groups or if the study lacked a control a group. Second, we applied a snowball technique whereby we reviewed the bibliographies of the selected five articles to identify potential additional works. This step resulted in the inclusion of three more articles, increasing the total number of articles that met our inclusion criteria to eight.

A possible reason for the relatively low number of articles is that research on vulnerable groups such as refugees is fraught with practical and ethical challenges. Practical challenges entail difficulties in accessing refugee camps, overcoming language barriers, and the generally high costs associated with conducting RCTs. When it comes to ethical barriers associated with RCTs, randomisation can be an issue. For example, when testing an intervention, one has to rely on sufficiently large numbers of children and youth with similar levels of needs/vulnerability within the population while also acknowledging that it is not possible to grant universal access to a specific intervention. Other ethical challenges relate to the risk of re-traumatisation, e.g. during survey interviews, which can (at least in part) be overcome through complementary qualitative approaches (Gaywood et al. 2020; Habib 2019).

The eight articles that were included in our shortlist for the final review are listed in Table 2. We collected information on the following variables for each article: spatio-temporal scope of intervention and target population; type of intervention; outcome of interest; research design; and main findings. Due to the variety of methodologies, sample sizes and intervention types, we provide qualitative discussion of these variables.

Table 2 Quantitative evidence on what works to improve access to education and quality of learning for refugee children
Full size table

Results

Although our search included articles published between 1990 and 2021, the eight studies that met our inclusion criteria were all published in or after 2007. The interventions took place in the United States, Canada, Lebanon, Germany, Ethiopia and Turkey. Only one study was conducted in a camp (in Ethiopia: Metzler et al. 2021), the others examined interventions implemented in public or community school settings.

Two of the eight studies examined interventions which were aimed at increasing access to education (de Hoop et al. 2019; Meloche et al. 2020), whereas the remaining six interventions were aimed at improving the quality of learning focusing either on second-language acquisition or on various academic outcomes (Busch et al. 2021; Krüger 2018; Metzler et al. 2021; Rousseau et al. 2007, 2014; Sirin et al. 2018). Sample sizes ranged from 62 (Krüger 2018) to 3,426 (Meloche et al. 2020). Some interventions examined multiple different outcomes such as academic outcomes and psychosocial well-being (Busch et al. 2021; Metzler et al. 2021; Rousseau et al. 2007, 2014), or both academic and non-academic outcomes (Meloche et al. 2020). The eight studies each examined different types of interventions: community school practices including 13 special services, cash transfer, language learning-enhanced physical education, enrolment in child and youth learning centres (CYLCs), an online game-based intervention, enrolment in an early childhood education programme, and two drama therapy workshops. Four out of the eight studies reported gender effects (de Hoop et al. 2019; Metzler et al. 2021; Rousseau et al. 2007, 2014). As for methodological sophistication, only one study applied a randomised control trial methodology (Rousseau et al. 2014); three applied a quasi-experimental design (Busch et al. 2021; de Hoop et al. 2019; Meloche et al. 2020); and four applied a pre- and post-test design (Krüger 2018; Metzler et al. 2021; Rousseau et al. 2007; Sirin et al. 2018). The target populations of all eight studies were refugee children aged between 3 and 18 years.

Interventions aimed at increasing refugees’ access to education

We identified only two articles which examined interventions aimed at improving refugees’ access to education (de Hoop et al. 2019; Meloche et al. 2020). In terms of evaluating enrolment, this low number is in itself worrying, given the high number of refugees who are deprived of education or drop out of school.

The two articles addressed different types of barriers to access, and both applied a quasi-experimental research design. Alysha Meloche et al. (2020) conducted their study in the United States, where they examined the different impacts of a school that implemented both community school (CS) and refugee centre programming compared to two other schools from the same district in terms of “academic, non-academic and college readiness outcomes of urban immigrant and refugee youth”, as they state in the title of their article (ibid.). The CS school’s refugee programme entailed 13 special services, such as citizenship classes, extended language support, mental health workshops and trauma-sensitive schooling, amongst others. The study further examined whether various academic and non-academic outcomes of English Language Learners (ELLs), a proxy for refugee status, differed from those of non-ELL students across the two school types (CS school and two comparison schools).

In this section, we report only on the results pertaining to non-academic outcomes, i.e. attendance rates and behavioural incidence (measured as out-of-school suspension rates). Meloche et al. (ibid.) found no statistical differences in attendance rates by ELL status or school type (each subgroup maintained an attendance rate of above 90 per cent over the period the study covered). The relatively low rates of dropouts however could not be causally connected to the CS practice given that there was a district-wide policy implemented to reduce absenteeism. The authors found that ELL students had more behavioural incidents than non-ELL students, however this was not present in the CS school. In fact, ELL students in the CS school were at the lowest risk of dropout, which is likely to be the result of the school counselling and trauma-sensitive training in place.

In the other study in this category, Jacobus de Hoop et al. (2019) evaluated an intervention addressing two critical barriers to access: the cost of transportation and foregone income because children are at school instead of working. The cash transfer programme (No lost Generation)Footnote 2 targeted Syrian displaced children in Lebanon. The intervention provided cash for the benefit of children enrolled in afternoon shifts at public primary schools and was designed to cover costs of commuting to school and to compensate the household for income foregone because children were at school instead of working. The authors applied a geographical discontinuity design to compare children in pilot governorates with children in neighbouring governorates and found no evidence that the cash transfer programme led to an increase in enrolment rates. However, children who received the programme benefits experienced increased household expenditures on education, and an increase in the probability of commuting to school by bus. Most importantly, children enrolled in afternoon shifts in the pilot areas spent approximately 20 per cent more time in school relative to children in the comparison areas. The results were similar across different age groups and between boys and girls.

Interventions aimed at improving the quality of education for refugees

We identified seven studies which examined interventions aimed at improving the quality of education for refugee children (Busch et al. 2021; Krüger 2018; Meloche et al. 2020; Metzler et al. 2021; Rousseau et al. 2007, 2014; Sirin et al. 2018). These included one experimental study (Rousseau et al. 2014), two quasi-experimental investigations (Busch et al. 2021; Meloche et al. 2020), and the remaining four used pre-and post-test design (Krüger 2018; Metzler et al. 2021; Rousseau et al. 2007; Sirin et al. 2018). Quality was measured in terms of learning outcomes, including literacy, numeracy and language skills.

One cluster randomised control trial examined the impact of a 12-week drama theatre and tutorship intervention programme in Canada on the maths and French grades of secondary school refugee students who were placed in special classes due to behavioural problems (Rousseau et al. 2014). The study found that for both maths and French grades, there was no significant difference in either the theatre or the tutorship intervention groups compared to the control group. It is interesting to note that an earlier study in our sample (Rousseau et al. 2007) found that a similar drama theatre intervention resulted in a significant improvement in mathematics for the experimental group compared to the control group, but no significant improvement was reported in either group with regard to French results. The main difference between the two studies was that the one conducted in 2007 focused on newly arrived refugees, while the 2014 randomised control trial examined refugee children who were first- and second-generation immigrants presenting with emotional and behavioural problems. This is an important difference between the two target populations since it shows that intervention effectiveness is likely to be conditioned by refugees’ length of stay in their host community.

One of the two quasi-experimental studies was the one by Meloche et al. (2020), already featured in the previous section above, which examined the different impacts on refugee students’ academic and non-academic outcomes by a school implementing community school (CS) and refugee centre programming compared to two other schools from the same district (ibid.). The study further examined whether various academic and non-academic outcomes of English Language Learners (ELLs), a proxy for refugee status, differed from those of non-ELL students across the two school types (CS school and two comparison schools). The authors found significant differences at all grade levels on overall grade point average (GPA) in each area (English, maths, science, social studies) at the end of year grades (p < .001). In sixth grade, non-ELL students attending the CS school outperformed all students in the comparison school in all measures. In seventh and eighth grades, both non-ELL and ELL students at the CS school scored significantly higher than non-ELL and ELL students in the comparison school across all measures. These results demonstrate that CS practices can benefit both refugee and non-refugee students’ learning outcomes.

The second of the two quasi-experimental studies we identified focused on the impact of a physical education (PE) intervention on second-language acquisition (Krüger 2018) in Germany. The intervention group received language-enriched PE sessions, while the control group did not receive any treatment. The authors tested refugee students’ domain-specific vocabulary learning, listening comprehension and use of prepositions. The intervention group showed better performance in the domain-specific vocabulary learning and listening comprehension tests, but no significant effect was found in case of the use of local prepositions (such as “upon”, “behind” etc.).

Another study which focused on second-language acquisition examined the impact of an online and game-based learning intervention (Cerego) for Syrian refugee children on Turkish language acquisition (Sirin et al. 2018).Footnote 3 In this particular study, learning through Cerego involved using 20 themed sets, each with 10–15 Turkish words and visual images. The authors found that students in the intervention group had significantly higher language scores than children in the control group.

The third study on language acquisition was carried out in Germany and examined the impact of an early childhood development programme (ECD) on newly arrived refugee children’s German language skills (Busch et al. 2021). The cross-sectional and longitudinal analyses found that attendance in the ECD was positively linked to improvements in German language skills, making ECD an important stage for second-language acquisition which can eventually facilitate a smoother transition to school.

The only study in our sample which was conducted in a camp setting examined the impact of child-and-youth-friendly spaces (CYLCs) at Buramino Camp in Dollo Ado, Ethiopia on educational, psychosocial and protection outcomes (Metzler et al. 2021). The study tested the impact of CYLC on literacy and numeracy skills of children enrolled in the programme between 3 and 6 months after the baseline data collection. The study found that younger children who attended CYLCs had major improvements in literacy and numeracy, although gains in numeracy were higher for boys than for girls. Older children who attended CYLCs recorded even bigger increases in both literacy and numeracy than younger ones, and older boys showed significantly greater improvements than girls both in literacy and numeracy.

Discussion

Despite being few in number, the studies we reviewed show that it is indeed possible to carry out rigorous quantitative evaluations in the context of forced displacement. Amongst the studies we reviewed, we found that cash transfer programmes can be effective in increasing the time refugee children spend in school by providing the financial means for families to send their kids to school. However, there was no evidence that cash transfers attracted a higher number of refugee children to school. Hence to increase enrolment, other types of intervention, such as targeted information campaigns and home visits by teachers, need to be evaluated.

There are some promising results regarding interventions to improve second-language acquisition, a major barrier both to access and quality learning. Evidence from our reviewed articles suggests that enrolment in early childhood development programmes can be appropriate for developing second-language skills and that in certain contexts, online, game-based solutions can also contribute to language acquisition. Child and youth learning centres in camp settings can improve refugee children’s numeracy and literacy skills, although more research is needed to find out how to improve these skills specifically for girls. Lastly, there was some evidence that theatre interventions can positively impact newly arrived refugees’ maths grades.

Despite these positive findings, the quantitative studies reviewed do suffer from some limitations regarding methodology, temporal aspects, target population and scope of analysis. Also, there is a lack of studies carried out in the Global South. Below, we discuss these limitations.

Limitations

Methodological limitations

The few studies that fit our inclusion criteria nonetheless exhibited some methodological limitations. The sample sizes were often small (Krüger 2018), there were often no control groups, only pre- and post-treatment evaluations (Metzler et al. 2021), and participants were often not randomly allocated to intervention and control groups (Busch et al. 2021).

Temporal limitations

Most studies evaluated the intervention shortly after the termination of the programme; hence we know little about the long-term effects. Furthermore, evaluations rarely considered the length of displacement which is however likely to influence the intervention’s impact (Rousseau et al. 2007, 2014).

Limitations regarding the target population

Amongst the eight articles we reviewed in detail, four reported gender-specific findings (de Hoop et al. 2019; Metzler et al. 2021; Rousseau et al. 2007, 2014). Results were often the same for boys and girls, but in one study, boys performed better than girls in both literacy and numeracy scores (boys also had higher baseline scores). This might have been caused by enrolment differences before displacement, gender differences in education engagement and variation in caregivers focus on girls’ learning (Metzler et al. 2021). This finding highlights the need for developing programmes specifically designed for helping refugee girls access quality education. There is also a need to conduct more research on internally displaced children. We did not find any study that addressed the wider environment of refugee children, i.e., teacher, parents and peers, even though these actors have been identified as important in quality learning (Burde et al. 2015).

Lack of comparison across intervention contexts

Studies were conducted in high-income countries (United States, Canada, Germany), one in an upper-middle-income country (Turkey), one in a lower-middle-income country (Lebanon), and only one study was conducted in a camp setting (Ethiopia). We did not find any study that compared the effectiveness of a particular intervention between low- and middle-income and high-income countries, hence we know little about the transferability of interventions to other contexts. Furthermore, there is a notable absence of studies which examine the situation of refugees displaced due to natural disasters; this gap has already been highlighted in earlier literature. Knowing what works regarding disaster risk reduction (DRR) education, or interventions that support children after disasters is urgently needed (Burde et al. 2015, 2017).

Lack of focus on individual project components

Studies usually examine interventions as a whole and rarely focus on specific components, making it difficult to understand which part of the intervention exactly influences educational outcomes. For example, the community and refugee school practice in the United States examined by Meloche et al. (2020) consisted of 13 special service components such as counselling, parent–teacher associations and cultural training, among others, but the research design did not envisage examining these components individually. This lack of disaggregation might mean that there were multiple pathways through which access to education or quality learning was impacted (equifinality) but we were unable to identify which project component was influential to what degree.

Conclusion

Even though millions of refugee children are in dire need of education, our systematic mapping of the literature indicates that there are only very few empirically rigorous quantitative studies that provide robust evidence on “what works” in improving these children’s access to education and quality learning. The eight studies which met our selection criteria suggest some promising results (e.g., cash transfers can increase educational attainment). However, since the studies are so few and varied with respect to geographical location, type of intervention and sample size, only limited conclusions can be drawn about the interventions’ general effectiveness and scalability.

Documenting the lack of such impact evaluations is an important finding in its own right, and one that demonstrates there is a significant need for future RCTs to provide information on the causal, direct impact of specific interventions. However, quantitative studies, when resources allow, should be combined with qualitative methods. Evaluations which combine these two research methods are the best suited to provide a comprehensive assessment of the effectiveness of a particular intervention.

Based on our scoping review, we have identified some areas for future research. First, as displacement has become increasingly protracted, in contexts where it is possible, it is important to compare a particular intervention’s impact across newly arrived refugees, refugees who have been staying in the host country/community for a longer period, and national/local citizens. Second, there is a need for more disaggregated data by sex, age groups, ethnicity, race, class, religious faith and disabilities. Knowing how the intersection of two or more of these attributes can promote or hinder access to education or quality learning is critical. Finally, it is crucial that interventions introduced during the COVID-19 pandemic-related school closures are evaluated. Remote or digital learning has often been unavailable for refugee children due to the lack of connectivity and necessary equipment. The increasing digital divide between different groups of learners can exacerbate existing barriers to access and quality.

Our systematic scoping review is an important step towards understanding the current refugee education intervention landscape. When it comes to the evaluation of interventions, we see two important structural challenges. First, while there is an increasing number of education interventions, most of them remain unevaluated. Our mapping of existing impact evaluations on what works to improve access to education and quality of learning for refugee children demonstrates that available rigorous quantitative evidence is scarce and often inconclusive. This is problematic because the cost-effectiveness or the overall performance of these programmes often remains unknown. Second, while the importance of multistakeholder engagement – researchers, practitioners and policymakers – is often highlighted (e.g. Siarova and van der Graaf 2022) in the field of refugee education, in reality significant gaps remain in interactions between these various actors. This in turn negatively impacts the development of sound programme design, and the collection and evaluation of empirical evidence on what works, when, and under what conditions in improving refugees’ access to and quality of education. Providing regular rigorous systematic reviews on refugee education interventions and fostering close collaboration between humanitarian and development actors, researchers and governments is crucial in improving refugee children’s lives and achieving SDG 4’s stipulated inclusive and equitable quality education for all by 2030.