Advertisement

20 Years of the Evidence Base on What Works to Prevent Child Marriage: A Systematic Review

Open AccessPublished:January 11, 2021DOI:https://doi.org/10.1016/j.jadohealth.2020.11.017

      Abstract

      Purpose

      This review assesses evaluations published from 2000 to 2019 to shed light on what approaches work, especially at scale and sustainably, to prevent child marriage in low- and middle-income countries.

      Methods

      We conducted a search of electronic databases and gray literature and evaluated the methodological quality and risk of bias of included studies.

      Results

      A total of 30 studies met the inclusion criteria. Interventions that support girls' schooling through cash or in-kind transfers show the clearest pattern of success in preventing child marriage, with 8 of 10 medium-high quality studies showing positive results. Although limited in number, five studies on favorable job markets and targeted life skills and livelihoods training show consistent positive results. Comparatively, asset or cash transfers conditional on delaying marriage show success only among two of four evaluations, and the three studies on unconditional cash transfers for poverty mitigation show no effect. Findings also show a low success rate for multicomponent interventions with positive results in only one of eight medium-high quality studies. Further, single component interventions were much more likely to be at scale and sustainable than multicomponent interventions.

      Conclusions

      These results indicate that enhancement of girls' own human capital and opportunities is the most compelling pathway to delaying marriage. In contrast, low rates of success, scale-up, and sustainability of multicomponent programs requires reconsideration of this approach.

      Keywords

      Implications and Contribution
      This review supports global efforts to meet the Sustainable Development Goal 5.3.1 target of ending child marriage by providing evidence on what works, highlighting the relative effectiveness, scale, and sustainability of interventions that enhance girls' human capital and opportunities through schooling, skills, and employment. There is not strong evidence for the effectiveness, scale or sustainability of multicomponent programs.
      See Related Editorial on p.833
      Child marriage, defined by the United Nations as a marriage or informal union under age 18 years and a manifestation of gender inequality, results in lifelong negative consequences for the health, well-being, and rights of millions of adolescent girls [
      • Santhya K.G.
      Early marriage and sexual and reproductive health vulnerabilities of young women: A synthesis of recent evidence from developing countries.
      ,
      • Marphatia A.A.
      • Ambale G.S.
      • Reid A.M.
      Women’s marriage age matters for public health: A review of the broader health and social implications in South Asia.
      ,
      • Koski A.
      • Clark S.
      • Nandi A.
      Has child marriage declined in Sub-Saharan Africa? An analysis of trends in 31 countries.
      ,
      • Olamijuwon E.O.
      • Chisumpa V.H.
      • Akinyemi J.O.
      Unveiling the realities of marrying too young: Implications of child marriage on sexual and reproductive health of girls and infant survival in Sub-Sahara Africa.
      ,
      • Godha D.
      • Gage A.J.
      • Hotchkiss D.R.
      • Cappa C.
      Predicting maternal health care use by age at marriage in multiple countries.
      ,
      • Delprato M.
      • Akyeampong K.
      The effect of early marriage timing on women’s and children’s health in Sub-Saharan Africa and Southwest Asia.
      ,
      UNICEF
      Child marriage [Internet].
      ]. It is closely associated with high rates of early pregnancies, maternal and child mortality or morbidity, and intimate partner violence during adolescence and with intergenerational poverty, poor health, and disempowerment for married girls and their children over the longer term [
      • Shapiro D.
      • Gebreselassie T.
      Marriage in Sub-Saharan Africa: Trends, determinants, and consequences.
      ,
      • Irani M.
      • Roudsari R.L.
      Reproductive and sexual health consequences of child marriage: A review of literature.
      ,
      • Raj A.
      When the mother is a child: The impact of child marriage on the health and human rights of girls.
      ]. Globally, the prevalence of child marriage among boys is just one sixth that among girls, with 5–20 times more girls married as children compared with boys in different countries. The range of sexual, reproductive, maternal, social, and economic consequences for girls are also much more severe than for boys [
      UNICEF
      Child marriage [Internet].
      ,
      Girls not brides
      SDGs and Child Marriage.
      ].
      Recent data indicate that one in five—or 12 million—girls are married as children annually, and 650 million women and girls currently alive were child brides [
      UNICEF
      Child marriage: Latest trends and future prospects.
      ,
      UNICEF
      Ending child marriage: Progress and prospects.
      ,
      • Malhotra A.
      • Elnakib S.
      Evolution in the evidence base on child marriage 2000-2019.
      ]. Although child marriages occur across the globe, 90% of the burden is in low- and middle-income countries (LMICs). The highest prevalence is in Sub-Saharan Africa (SSA) at 37%, followed by South Asia at 30%, but South Asia is home to the largest number of child brides, a distinction that will soon belong to SSA with a growing population and slower declines in child marriage rates [
      UNICEF
      Child marriage [Internet].
      ].
      In 2016, the elimination of child marriage by 2030 became an international commitment under goal 5 on gender equality in the Sustainable Development Goals (SDGs), leading to a rapid increase in policy and programmatic efforts to end the practice. Data, tracking, and accountability mechanisms for measuring progress on the indicator for SDG target 5.3.1--the proportion of women aged 20–24 years married or in union before age 18 years-- have also multiplied [
      United Nations
      Goal 5: Sustainable development knowledge platform [Internet]. Sustainable development goals.
      ]. Despite this expansion, there is a well-acknowledged concern that progress to match the scope of the problem remains elusive. A 2018 trend analysis by United Nations Children's Fund (UNICEF) indicates that while globally the rate of child marriage has declined from 25% to 21% in the last decade, many high prevalence countries have seen stagnation, and even increases. Dauntingly, progress would have to increase 12-fold in the remaining 10 years of the SDGs to achieve target 5.3.1 by 2030 [
      UNICEF
      Ending child marriage: Progress and prospects.
      ].
      This monumental task puts pressure on the global community to deliver on effective interventions and policies at scale, generating strong interest in a more systematic understanding of both the effectiveness and reach of child marriage programs [
      • Finnie A.
      • Malhotra A.
      • Travers E.
      • et al.
      Advancing the evidence base on strategies to end child marriage & support married girls. Conference Proceedings. 2019 Sept 30-Oct 2; Geneva, Switzerland.
      ,
      • Svanemyr J.
      • Chandra-Mouli V.
      • Raj A.
      • et al.
      Research priorities on ending child marriage and supporting married girls.
      ]. Recent reviews and convenings on the issue have especially articulated the need to understand whether the many comprehensive or “multicomponent” programs addressing child marriage are resulting in later marriages among significantly large numbers of girls or whether some of the more targeted or “single-component” interventions are more effective and expansive. A 2018–2019 evaluation and review of the first phase of the Global Programme on Child Marriage coordinated by UNICEF and the United Nations Fund for Population Activities (UNFPA) articulated this as a priority concern for moving forward with its Phase 2 efforts. The question also emerged as a central theme at a 2019 international convening of experts by World Health Organization, UNICEF, and Girls Not Brides and is echoed in several recent policy and research reviews on child marriage [
      Girls not brides
      SDGs and Child Marriage.
      ,
      • Finnie A.
      • Malhotra A.
      • Travers E.
      • et al.
      Advancing the evidence base on strategies to end child marriage & support married girls. Conference Proceedings. 2019 Sept 30-Oct 2; Geneva, Switzerland.
      ,
      • Forte C.L.
      • Plesons M.
      • Branson M.
      • Chandra-Mouli V.
      What can the global movement to end child marriage learn from the implementation of other multi-sectoral initiatives?.
      ].
      Clearer evidence on the effectiveness of multicomponent versus specific single-component interventions is critically important for policy makers, donors, and advocates, as they struggle to urgently and efficiently support governments in realizing commitments to end child marriage within the SDG time frame, especially through the implementation of National Action Plans on child marriage adopted by a growing number of countries in the last 5 years [
      Lessons learned from national initiatives to end child marriage - 2016 [Internet]. Girls Not Brides.
      ]. The resources, capacity, alliances, and coordination required to implement multicomponent interventions spanning a range of sectors and ministries are likely to be very different from those required to implement specific single-component interventions with fewer stakeholders, but possibly different challenges [
      • Forte C.L.
      • Plesons M.
      • Branson M.
      • Chandra-Mouli V.
      What can the global movement to end child marriage learn from the implementation of other multi-sectoral initiatives?.
      ].
      Lack of clarity on which interventions are effective is in part because of the limited evidence base available from previous analyses. Historically, evaluations of programs with child marriage prevention as an outcome have been limited in number and quality, with a significant increase only after 2015 [
      • Jain S.
      • Kurz K.
      New insights on preventing child marriage: A global analysis of factors and programs.
      ,
      • Das Gupta S.
      • Pande R.
      Knot ready: Documenting initiatives to delay early marriage: Update 1.
      ]. Two earlier systematic reviews have contributed substantially to our understanding, but although both confirmed that programmatic efforts can delay marriage, both also concluded with equivocal findings regarding the success rates of any given type of intervention [
      • Kalamar A.M.
      • Lee-Rife S.
      • Hindin M.J.
      Interventions to prevent child marriage among young people in low- and middle-income countries: A systematic review of the published and gray literature.
      ,
      • Lee-Rife S.
      • Malhotra A.
      • Warner A.
      • Glinski A.M.
      What works to prevent child marriage: A review of the evidence.
      ]. In their 2012 systematic review, Lee-Rife et al. used relatively generous selection criteria to examine 23 evaluations published between 1991 and 2010, 12 of which were low quality. They concluded that although both horizontal (or multicomponent) as well as more focused vertical (or single-component) interventions showed some promise, neither category demonstrated clear positive results [
      • Lee-Rife S.
      • Malhotra A.
      • Warner A.
      • Glinski A.M.
      What works to prevent child marriage: A review of the evidence.
      ]. The second review in 2016 by Kalamar et al. used stricter selection criteria, thus limiting itself to only 11 higher quality evaluations from 2000 to 2015, which permitted sparse representation for any given intervention category. They concluded that every type of intervention had positive, negative, and mixed results [
      • Kalamar A.M.
      • Lee-Rife S.
      • Hindin M.J.
      Interventions to prevent child marriage among young people in low- and middle-income countries: A systematic review of the published and gray literature.
      ].
      A second reason for the gap between the existing evidence and the expressed needs of the field is that few impact evaluations or evidence reviews have incorporated a focus on scale and sustainability [
      • Kalamar A.M.
      • Lee-Rife S.
      • Hindin M.J.
      Interventions to prevent child marriage among young people in low- and middle-income countries: A systematic review of the published and gray literature.
      ,
      • Lee-Rife S.
      • Malhotra A.
      • Warner A.
      • Glinski A.M.
      What works to prevent child marriage: A review of the evidence.
      ,
      • Chae S.
      • Ngo T.
      The global state of evidence on interventions to prevent child marriage.
      ]. However, as an estimated 120 million girls are at risk of child marriage over the next decade, policy makers are looking for recommendations on programs that are not only effective but can consistently reach hundreds of thousands or millions of girls in the most affected countries. Given the wide variation in population size across countries, even if we define scale conservatively as reaching 25% of the population at risk (still far from the target of elimination), a rough estimation shows us that a smaller country such as Zambia with approximately 500,000 girls at risk over the next 10 years would have to reach at least 125,000 girls, whereas a larger country such as Bangladesh with approximately 9 million girls at risk would have to reach at least 2.25 million girls.
      A third challenge to clearer, more actionable recommendations from the existing evidence base has been the ambiguity and inconsistency in the conceptual framing of child marriage programs, making it difficult to classify interventions in discreet categories. In particular, there is only an implicit distinction between three common approaches to child marriage prevention—personal empowerment, social norm change, and structural shifts—because several interventions overlap across these approaches and have multiple pathways to change. For example, a significant share of programs aims to prevent child marriage by empowering girls while also engaging families and communities for social norm change [
      • Ghimire A.
      • Samuels F.
      Continuity and change in social norms and practices around marriage and education in Nepal. London Overseas Development Institute 2014.
      ,
      • Steinhaus M.
      • Hinson L.
      • Rizzo A.T.
      • Gregowski A.
      Measuring social norms related to child marriage among adult decision-makers of young girls in Phalombe and Thyolo, Malawi.
      ,
      • Cislaghi B.
      • Mackie G.
      • Nkwi P.
      • Shakya H.
      Social norms and child marriage in Cameroon: An application of the theory of normative spectrum.
      ,
      • Kenny L.
      • Koshin H.
      • Sulaiman M.
      • Cislaghi B.
      Adolescent-led marriage in Somaliland and Puntland: A surprising interaction of agency and social norms.
      ]. They typically include some combination of exposure to information, skills, social networks, safe spaces, etc. for the girls so that they are more capable, have more self-confidence, and are able to negotiate later marriages. Generally, empowerment approaches are delivered in conjunction with communication or mobilization campaigns to change family and community attitudes toward child marriage [
      • Mathur S.
      • Mehta M.
      • Malhotra A.
      Youth reproductive health in Nepal: Is participation the answer?.
      ,
      • Amin S.
      Empowering adolescent girls in rural Bangladesh: Kishori Abhijan.
      ,
      • Levitt-Dayal M.
      • Motihar R.
      • Kanani S.
      • Mishra A.
      Adolescent girls in India choose a better future: An impact assessment of an educational programme.
      ,
      • Erulkar A.S.
      • Muthengi E.
      Evaluation of Berhane Hewan: A program to delay child marriage in rural Ethiopia.
      ]. However, several of these programs also include access to education or job skills as part of personal empowerment strategies. And yet, programs focusing more exclusively on schooling or employment would generally be considered structural interventions that increase girls' human capital or opportunities, by making schooling or work viable alternatives to marriage—a very different classification and pathway to change.
      Of course, with a large literature documenting the strong and consistent inverse relationship between girls' education and child marriage across almost every context, it is likely that education—especially at the secondary level—could delay marriage through multiple pathways. Being in school could simply preclude marriage as schooling and marriage are seen as mutually exclusive in most societies. Another path could be through girls acquiring skills, confidence, friends, and nondomestic options and opportunities. A critical mass of girls going to school could also result in social norm change, as families start considering adolescent school girls as children rather than as marriageable adults [
      • Glick P.
      • Handy C.
      • Sahn D.E.
      Schooling, marriage, and age at first birth in Madagascar.
      ,
      • Adu Boahen E.
      • Yamauchi C.
      The effect of female education on adolescent fertility and early marriage: Evidence from free compulsory universal basic education in Ghana.
      ,
      • Rasmussen B.
      • Maharaj N.
      • Sheehan P.
      • Friedman H.S.
      Evaluating the employment benefits of education and targeted interventions to reduce child marriage.
      ,
      • Wodon Q.T.
      • Jesse C.
      • Belhaj N.
      • et al.
      Tanzania economic update: The power of investing in girls-Educating girls and ending child marriage in Tanzania.
      ,
      • Malhotra A.
      • Amin A.
      • Nanda P.
      Catalyzing gender norm change for adolescent sexual and reproductive health: Investing in interventions for structural change.
      ]. Thus, education has the potential to be simultaneously a structural, empowerment, and norm change intervention, its promise often depending on the level and quality of schooling. It is also important to note that most evaluated interventions on education's effect on delaying marriage have been limited to cash transfers or other subsidies for school attendance, thus aiming to increase demand for education among girls and their families [
      • Erulkar A.
      • Medhin G.
      • Weissman E.
      The impact and cost of child marriage prevention in rural Ethiopia.
      ]. Supply-side education interventions—such as more secondary schools, female teachers, better transport, or curricula—remain largely untested for their impact on child marriage prevention [
      Girls not brides
      SDGs and Child Marriage.
      ].
      Other structural interventions may also operate through multiple pathways, making it difficult to classify them in only one category. For example, asset transfers as a reward for girls marrying later may structurally improve economic options while normatively also increasing the value of unmarried daughters to families. Thus, a single intervention working through multiple pathways could potentially have as much or even stronger impact compared with multiple interventions that work through single pathways. From a policy and implementation perspective, moreover, the ambiguous theoretical distinctions between intervention approaches may be both confusing and of secondary importance compared with clarity on each intervention's content and effectiveness.
      It is with the aim of providing greater clarity for policymakers and implementers on the effectiveness of specific interventions that in this article, we undertake a systematic review of evaluations on child marriage prevention in LMICs published between 2000 and 2019. By incorporating 16 additional evaluated interventions —13 published since 2016—we are able to go beyond previous reviews and cover a total of 30 evaluations, allowing us to undertake a more fine-tuned analysis of intervention categories and their effectiveness. Several of these are multiarm studies that examine the comparative impact of different interventions. Thus, we are able to unpack and classify single- and multi-component interventions in more nuanced rather than broad categories. In addition, we include scale and sustainability of interventions as key dimensions in our analysis to make the findings more directly relevant to the expressed needs of program and policy actors. Overall, our findings move the field closer to a clearer understanding of whether specific multicomponent or single-component interventions have been more successful in preventing child marriage sustainably, and at scale.
      The main research question guiding our review is, “What interventions are effective in delaying age at marriage or reducing child marriage incidence among girls aged 10–24 years as assessed in experimental and quasi-experimental studies?”

      Methods

      Data sources

      This review builds on a comprehensive scoping review of studies published between January 2000 and July 2019 and generating evidence on child marriage in LMICs, from which we focus on only those studies that evaluated the impact of interventions to prevent child marriage. As the literature--especially before 2015--often refers to child marriage and early marriage interchangeably, we follow previous systematic reviews in including studies on child or early marriage prevention. A prespecified research protocol that describes the review aims and methods for both the published and gray literature was followed and is available on request from the authors. We developed a broad search strategy to maintain breadth of coverage, using a combination of keywords to search a total of eight electronic databases: PubMed, EMBASE, Cochrane, PsycINFO, CINAHL, Popline, Sociological Abstracts, and Econlit. The PubMed search strategy is presented as follows:
      ((Child(tiab) AND Marriage∗[tiab]) OR (adolescen∗[tiab] AND marriage∗[tiab])OR (Early(tiab) AND Marriage∗[tiab]) OR “Child Marriage”[tiab] OR “Early Marriage”[tiab] OR “Forced marriage” [tiab] OR “child bride” [tiab] OR “Adolescent marriage"[tiab]) AND (“2000”[PDAT]: “2019”[PDAT])
      To ensure comprehensiveness of the search, we additionally mapped gray literature through a web-based search of UN and nongovernmental organizations (NGOs) engaged in research or interventions on child marriage. We also handsearched the literature based on identified citations in the published and gray literature for additional titles. We followed the PRISMA guidelines for the reporting of systematic reviews and meta-analyses.

      Inclusion/exclusion criteria

      Articles were included if the following criteria were met: (1) article evaluated the impact of one or more interventions to prevent child marriage or delay marriage among 10- to 24-year-old girls; (2) evaluation study included at least one quantitative behavioral measure of the impact of the intervention(s) on child marriage prevention. Quantitative measures needed to capture child marriage by either examining entry into marriage among girls under age 18 years or by examining marriage before the age of 18 years among older women (such as women aged 20–24 years); (3) evaluation used an experimental or quasi-experimental evaluation design; (4) article was published in the English language; and (5) article was published between January 1, 2000, and July 30, 2019.

      Data extraction

      The search results were exported into Covidence and Endnote reference manager software, and duplicates were removed. Given the large number of studies, titles and abstracts were screened by one of the authors in close and frequent consultation with the other author, and candidate articles for full-text review were reviewed by both authors to ascertain their eligibility for inclusion. Disagreements about inclusion were resolved through discussion between the two reviewers. Data extraction was carried out by both reviewers using a standardized template that included the following domains: intervention and evaluation dates, intervention objective, intervention content including whether the intervention was multicomponent or single component or multiarm, study design, sampling method, sample size, analysis, attrition, and key findings (Data extraction template available in Supplemental Table 2). Detailed information on the outcome was extracted, and when available, adjusted estimates were reported instead of crude estimates. Where studies lacked necessary details, we made an effort to contact authors, requesting additional information. However, this was not possible for all studies, given that many publications described interventions that took place many years before the date of publication.

      Quality assessment

      We conducted a quality assessment that included both an assessment of methodological quality and risk of bias. Building on an existing risk of bias tool [
      • Kennedy C.E.
      • Fonner V.A.
      • Armstrong K.A.
      • et al.
      The evidence project risk of bias tool: Assessing study rigor for both randomized and non-randomized intervention studies.
      ], we assessed the quality of included studies using a scoring system based on the following domains: (1) study design, attrition, and sample size; (2) selection bias, measurement of exposure, spillover/contamination; (3) estimation techniques and confounding; and (4) outcome measurement and clarity of reporting. Studies were given a score ranging from 1 to 12 and were subsequently classified into low (score below 7), medium (score of seven–9) and high (score above 9) quality. The two authors independently assessed the quality of each study. Where discrepancies or conflicts arose, the decision was resolved through discussion among the reviewers. Supplemental Table 1 outlines the results of the assessment.

      Analysis

      We analyzed the included studies on both intervention and evaluation characteristics:
      • 1.
        The intervention location, implementer(s), period, purpose, content, reach, and sustainability.
      • 2.
        The evaluation publication year, evaluator(s), study period, design, methodology, quality, outcome measure(s) on child marriage prevention, results, and direction of findings. Although most studies included other behavioral and attitudinal outcomes--such as school retention, pregnancy prevention, self-efficacy, negotiation, etc.--lack of comparability in such outcomes across studies precluded an analysis of any additional outcomes.
      As the variation in outcome measures on child marriage reduction used across studies did not permit the assessment of study results using a common measure, we analyzed findings on whether programs showed a positive, mixed, null, or negative effect. We considered a study as demonstrating a positive impact if its results found the intervention to successfully prevent early or child marriage and were statistically significant (p < .05). In addition, we conducted a sensitivity analysis excluding studies that were rated as low quality and noted the effects this had on the overall findings. In view of the heterogeneity in intervention design and delivery—specifically with respect to intervention duration and intensity—as well as participant characteristics and diversity in outcome measures, it was not possible to undertake a full meta-analysis.

      Results

      Study selection

      Our search yielded a total of 3,556 records: 3,254 from electronic databases and 302 from handsearching and a scan of the gray literature. After removing 1,267 duplicates, 1,987 publications were screened on title and abstract, of which 61 articles were retained for full-text assessment. The reasons for excluding full-text articles are listed in Figure 1. In total, we included 34 studies published between 2000 and 2019, covering 30 distinct evaluations of child marriage prevention programs implemented in LMICs.

      Study characteristics

      Table 1 provides a summary description of the characteristics of the included studies, whereas Table 2 provides more specifics on the interventions. Both tables are organized first by the two broad categories of interest: multicomponent (11 studies) and single component (including 10 program interventions and two macro policies) (12 studies). They also include a third category of seven multiarm studies, which evaluated both multi- and single-component interventions. It is noteworthy that these studies—published largely since 2015—were undertaken with a deliberate interest in untangling the potential unbundled impact of bundled interventions. Single-component evaluations are also weighted toward more recent publication (7/12 since 2014). In contrast, a significant majority of multicomponent studies (8/11) were published from 2000 to 2012, reflecting the early wave of interest in comprehensive child marriage interventions. Earlier studies are also more likely to cover South Asia, whereas later studies have a greater focus on SSA, with the growing concerns about high rates and burden in that region.
      Table 1Evaluation studies with child or early marriage prevention as an outcome, 2000–2019; Key characteristics by type of intervention: multicomponent, single component, and multiarm
      Study and locationIntervention reach and participantsSustainabilityEvaluation designEvaluation qualityEvaluation CM outcome measure(s)Results
      Multicomponent
       CEDPA, 2001, India [
      Centre for Population and Development Activities
      Adolescent girls in India choose a better future: An impact assessment.
      ]
      10,000

      F, 12–24 years
      Not sustainedQuasi-experimental post-only retro dataLowRelative risk of marriage <18 years

      F, 12–24 years
      Positive

      RR comparing nonparticipants to participants = 1.35
       Mathur, Mehta, & Malhotra, 2004, Nepal [
      • Mathur S.
      • Mehta M.
      • Malhotra A.
      Youth reproductive health in Nepal: Is participation the answer?.
      ]
      <1,000

      F and M, 14–21 years
      Not sustainedQuasi-experimental pre-post case and controlLowProportion F 14–21 who marriedMixed

      Significant difference between participants and nonparticipants in urban areas (23.6% vs. 14.6%)

      NS difference in prop in rural areas
       Kanesnathan et al., 2008, India [
      • Kanesathasan A.
      • Cardinal L.J.
      • Pearson E.
      • et al.
      Catalyzing change: Improving youth sexual and reproductive health through DISHA, an integrated program in India.
      ]
      12,000

      F and M 14–24 years
      Not sustainedQuasi-experimental pre-post case and controlHighPercent F 14–24 married <18 years; mean age at marriageNull

      60% married <18 years at baseline compared with 40% at endline (NS).

      Mean of 15.9 at baseline vs. 17.8 at endline (NS).
       Shahnaz et al., 2008, Bangladesh [
      • Shahnaz R.
      • Karim R.
      Providing microfinance and social space to empower adolescent girls: An evaluation of BRASC’s ELA Centres.
      ]
      30,000–40,000

      Female 10–24 years
      SustainedQuasi-experimental pre-post case and controlLowOdds of marriage F 10–22 in 2 years from baseline to endlinePositive

      Logit coefficient = −.385 (participants vs. nonparticipants)
       UNICEF, 2008; Diop, Moreau, Benga, 2008, Senegal [
      • Diop N.J.
      • Moreau A.
      • Benga H.
      Evaluation of the long-term impact of the TOSTAN programme on the abandonment of FGM/C and early marriage: Results from a qualitative study in Senegal.
      ,
      UNICEF
      Long-term evaluation of the Tostan programme in Senegal: Kolda, Thiès and Fatick regions.
      ]
      6,000–9,000 F and M adultsNot sustainedQuasi-experimental post onlyLowPercent F 20–24 married <18 yearsNull

      81% married <18 years in trt villages compared with 80% in control.
       Erulkar and Muthengi, 2009, Ethiopia [
      • Erulkar A.S.
      • Muthengi E.
      Evaluation of Berhane Hewan: A program to delay child marriage in rural Ethiopia.
      ]
      420

      F, 10–19 years
      Not sustainedQuasi-experimental pre-post case and controlMediumHazard of marriage F 10–14 and 15–19 married in 2 years from baseline to endlineMixed

      10–14 years: HR comparing participants to nonparticipants = .1 (Sig)

      15–19 years: HR = 2.4 (Sig)
       Amin et al., 2011, Bangladesh [
      • Amin S.
      Empowering adolescent girls in rural Bangladesh: Kishori Abhijan.
      ]
      15,000

      F 13–22 years
      Not sustainedQuasi-experimental

      Pre-post and Matched participants, nonparticipants
      LowOdds of marriage

      F 13–22 in 2 years from baseline to endline
      Null

      OR comparing participants to nonparticipants = 1.04 (NS)
       Daniel & Nanda, 2012, India [
      • Daniel E.E.
      • Nanda R.
      The effect of reproductive health communication interventions on age at marriage and first birth in rural Bihar, India: a retrospective study. Watertown: Pathfinder International.
      ]
      17,000

      F 15–19 years
      UnknownQuasi-experimental post onlyLowHazard of marriage F 15–19 married in 5 years from program to endline; median [email protected]Positive

      HR comparing participants and nonparticipants = .56 (Sig);

      Median age of marriage 22 years among participants vs. 19.4 nonpartcipants (Sig)
       Stark et al., 2018, Ethiopia [
      • Stark L.
      • Asghar K.
      • Seff I.
      • et al.
      Preventing violence against refugee adolescent girls: Findings from a cluster randomised controlled trial in Ethiopia.
      ]
      986

      F 13–19 years
      Not sustainedCluster RCT treatment and controlLowOdds of marriage

      F 13–19 years during 1 year of intervention
      Null

      OR comparing participants and nonparticipants = .72 (NS)
       Stark et al., 2018, DRC [
      • Stark L.
      • Seff I.
      • Asghar K.
      • et al.
      Building caregivers’ emotional, parental and social support skills to prevent violence against adolescent girls: Findings from a cluster randomised controlled trial in Democratic Republic of Congo.
      ]
      869

      F 10–14 years
      Not sustainedCluster RCT treatment and controlMediumOdds of marriage

      F 13–14 years during 1 year of intervention
      Null

      OR comparing participants and nonparticipants = 1.24 (NS)
       Bandiera et al., 2018, Uganda [
      • Bandiera O.
      • Buehren N.
      • Burgess R.
      • et al.
      Women’s empowerment in action: Evidence from a randomized control trial in Africa.
      ]
      50,000

      F 14–20 years
      SustainedCluster RCT treatment and controlMediumLikelihood of marriage F 14–20 years married in 2 and 4 years of interventionPositive

      Midline: 6.9% less likely to be married

      Endline: 8% less likely to be married
      Single component
       Angrist et al., 2003, Colombia [
      • Angrist J.
      • Bettinger E.
      • Bloom E.
      • et al.
      Vouchers for private schooling in Colombia: Evidence from a randomized natural experiment.
      ]

      125,000

      F and M 13–17 years
      UnknownQuasi-experimental lottery winners and losersMediumProbability 13–17 years F and M marry in 1- to 4-year program exposurePositive

      Participants 1.1 pp less likely to be married or cohabitating compared with nonparticipants (Sig)
       Pande et al., 2006, India [
      • Pande R.
      • Kurz K.
      • Walia S.
      • et al.
      Improving the reproductive health of married and unmarried youth in India.
      ]
      440, F 12–19 yearsUnknownQuasi-experimental pre-post case and controlMediumMedian age at marriage; Percent F 12–19 years marrying <18 years;Positive

      1-year increase in median age in trt compared with no change in control

      18.9% decrease in marriage <18 years in trt versus no significant decrease in control
       Gulemetova, 2011, Mexico [
      • Gulemetova M.
      Evaluating the impact of conditional cash transfer programs on adolescent decisions about marriage and fertility: The case of Oportunidades.
      ]
      Millions

      M and F school age
      SustainedQuasi-experimental treatment versus in waitingHighHazard of marriage F 13–19 years marry in 2- to 8-year program exposurePositive

      Instantaneous hazard rate of marriage lowered (hazard coefficient = −.76 Sig)
       Alam, Baez, Del Carpio, 2011, Pakistan [
      • Alam A.
      • Baez J.E.
      • Del Carpio X.V.
      Does cash for school influence young women’s behavior in the longer term? Evidence from Pakistan.
      ]
      150,000

      Middle school age girls
      SustainedQuasi-experimental treatment versus controlHighProbability F 15–19 years married; up to 4-year program exposure; age at marriagePositive

      No effect on probability of marriage, but delayed age at marriage by 1.2–1.5 years (Sig)
       Jensen, 2012, India [
      • Jensen R.
      Do labor market opportunities affect young women’s work and family decisions? Experimental evidence from India.
      ]
      4,000

      F 15–21 years
      SustainedCluster RCT

      160 study and control villages
      HighLikelihood of marriage F 15–21 years in 3 years of interventionPositive

      5.1% lower likelihood of marriage for program participants (Sig)
       Heath and Mobarek, 2014, Bangladesh [
      • Heath R.
      • Mobarak A.M.
      Manufacturing growth and the lives of Bangladeshi women.
      ]
      2+ million young female workersSustainedQuasi-experimental

      60 study and control villages
      HighHazard of marriage girls 12–18 years up to 7-year exposurePositive

      A 28% lower hazard of marriage for girls in garment proximate villages
       Hallfors et al., 2015, Zimbabwe [
      • Hallfors D.D.
      • Cho H.
      • Rusakaniko S.
      • et al.
      The impact of school subsidies on HIV-related outcomes among adolescent female orphans.
      ]
      328

      F avg 12 years in Grade 6
      UnknownCluster RCT

      25 primary schools; five waves
      MediumOdds of marriage F average age 12 in 5-year program exposurePositive

      Odds of marriage was .37 lower for intervention participants compared with controls after 5 years
       Handa et al., 2015, Kenya [
      • Handa S.
      • Peterman A.
      • Huang C.
      • et al.
      Impact of the Kenya cash transfer for orphans and vulnerable children on early pregnancy and marriage of adolescent girls.
      ]

      150,000

      Poor households
      UnknownCluster RCT

      28 treatment vs. in waiting locations
      HighLikelihood of marriage

      F 12–24 years in 1- to 4-year program exposure
      Null

      Girls 2 pp less likely to marry in treatment group (NS).
       Nanda et al., 2016, India [
      • Nanda P.
      • Datta N.
      • Das P.
      • et al.
      Making change with cash? Impact of a conditional cash transfer program on age of marriage in India.
      ]
      50,000

      F born

      1994–1997
      SustainedQuasi-experimental

      Matched beneficiaries and nonbeneficiaries—post only
      MediumProbability of F 18 being married and marrying before 18 yearsNull

      Proportion of girls married <18 years not different between participants and nonparticipants. (probit = .61 NS)
       Dake et al., 2018, Malawi and Zambia [
      • Dake F.
      • Natali L.
      • Angeles G.
      • et al.
      Cash transfers, early marriage, and fertility in Malawi and Zambia.
      ]
      380,000+ poor householdsSustainedCluster RCT

      29 clusters in Malawi, 92 Zambia
      HighProbability F and M 14–21 years marry in 2- to 3-year program exposureNull

      No effect on risk of ever marriage in Malawi (−.00428 NS) or Zambia (.0117 NS).
       Koski et al., 2018, 8 African countries [
      • Koski A.
      • Strumpf E.C.
      • Kaufman J.S.
      • et al.
      The impact of eliminating primary school tuition fees on child marriage in Sub-Saharan Africa: A quasi-experimental evaluation of policy changes in 8 countries.
      ]
      Million+

      F and M primary school children
      SustainedQuasi-experimental

      Cohorts pre and post fee elimination
      MediumOdds of marriage <15 and <18 years

      F 15–49 years w/prog exposure
      Mixed

      Sig 2% decline in marriage <15 years; Not Sig decline marriage

      < 18 - variable effect by country
       Hahn et al., 2018, Bangladesh [
      • Hahn Y.
      • Islam A.
      • Nuzhat K.
      • et al.
      Education, marriage, and fertility: Long-term evidence from a female stipend program in Bangladesh.
      ]
      Million+

      F in grades 6–10


      Sustained
      Quasi-experimental

      Cohorts before and after stipend
      HighDelay in entry to 1st marriage—F 6–1

      4 years as they age to 23–31 w/2 and 5-year program exposure
      Positive

      5-year stipend: first marriage delayed by .57 years; 2-year stipend, first marriage delayed by .34 years
      Multiarm
       Duflo, Dupas, and Kremer, 2015, Kenya [
      • Duflo E.
      • Dupas P.
      • Kremer M.
      Education, HIV, and early fertility: Experimental evidence from Kenya.
      ]
      19,300

      F and M

      13–20 years
      Not sustainedCluster RCT

      328 schools randomized into three arms
      HighLikelihood of marriage

      M and F 13–20 years in 3, 5, 7 years of program
      Mixed by arm

      At 3 years:

      CCT-schl: 2.6pp less likely to be married (Sig)

      HIV ed: Null

      Multicomp: Null

       Baird, McIntosh, and Ozler, 2016, 2011, 2009, Malawi [
      • Baird S.
      • McIntosh C.
      • Özler B.
      When the money runs out: Do cash transfers have sustained effects on human capital accumulation?.
      ,
      • Baird S.
      • Chirwa E.
      • McIntosh C.
      • Özler B.
      The short-term impacts of a schooling conditional cash transfer program on the sexual behavior of young women.
      ,
      • Baird S.
      • McIntosh C.
      • Özler B.
      Cash or condition? Evidence from a cash transfer experiment.
      1,000–2,000

      F 13–22 years
      Not sustainedCluster RCT

      88 treatment and 88 control areas
      HighLikelihood of marriage

      F 13–22 in two program exp and 2-year postprogram
      Mixed by arm

      CCT school: 11%–16% less likely to be ever married whereas 2 years earlier there had been no effect.

      UCT: earlier effect to delay marriage is gone.
       Amin et al., 2016; 2018 Bangladesh [
      • Amin S.
      • Ahmed J.
      • Saha J.
      • et al.
      Delaying child marriage through community-based skills-development programs for girls: Results from a randomized controlled study in rural Bangladesh.
      ,
      • Amin S.
      • Saha J.S.
      • Ahmed J.A.
      Skills-building programs to reduce child marriage in Bangladesh: A randomized controlled trial.
      ]
      11,609

      F 12–18 years
      Not sustainedCluster RCT

      72 intervention and 24 control communities
      HighHazard of marriage <18 years at 18-month intervention

      F 12–18 years
      Positive all arms

      Adjusted HR of marriage <18 decreased in all arms about equally (.70 livelihoods; .72 gender; .75 education) program impact stronger for <16
       Erulkar, Medhin & Weissman, 2017, Ethiopia [
      • Erulkar A.
      • Medhin G.
      • Weissman E.
      The impact and cost of child marriage prevention in rural Ethiopia.
      ]
      <10,000

      F 12–17 years
      Not sustainedQuasi-experimental pre and post but no controlLowRisk of ever marriage girls 12–14 and 15–17 in 28 months of programMixed all arms

      For 12–14

      Community dialog RR: .42 (Sig)

      Education arm RR: .09 (Sig)

      For 15–17

      CCT Asset RR: .57 (Sig)

      Comprehensive arm RR: .38 (Sig)

      Other arms null.
       Erulkar et al., 2017, Tanzania [
      • Erulkar A.
      • Medhin G.
      • Weissman E.
      The impact and cost of child marriage prevention in rural Tanzania.
      ]

      <10,000

      F 12–17 years
      Not sustainedQuasi-experimental pre-post case and controlMediumRisk of ever marriage girls 12–14 and 15–17 in 28 months of programMixed by arm

      For 12–14

      Multicomp RR: .33 (Sig)

      For 15–19

      CCT asset RR: .52 (Sig)

      All other arms null.
       Erulkar et al., 2017, Burkina Faso [
      • Erulkar A.
      • Medhin G.
      • Weissman E.
      The impact and cost of child marriage prevention in rural Burkina Faso.
      ]
      <10,000

      12–17 F years
      Not sustainedQuasi-experimental pre-post case and controlMediumRisk of ever marriage girls 15–17 in 28 months of programMixed by arm

      For 15–17

      Community dialog RR: .33 (Sig)

      All other arms null
       Buchmann et al., 2018, Bangladesh [
      • Buchmann N.
      • Field E.
      • Glennerster R.
      • et al.
      Power vs money: Alternative approaches to reducing child marriage in Bangladesh, a randomized control trial.
      ]
      46,000

      F 10–19 years
      UnknownCluster RCT

      460 communities randomized into four arms
      HighLikelihood of marriage <18 years girls 15–17 in 2.5-year program exposure and 4.5 post programMixed by arm

      CCT incentive: girls 21% less likely to be married <18 years (Sig)

      Empowerment: null

      Comprehensive: null
      CCT = conditional cash transfer; F = female; HR = hazard ratio; M = male; NS = nonsignificant; OR = odds ratio; RCT = randomized controlled trial; RR = risk ratio; Sig = significant; UCT unconditional cash transfer.
      Table 2Specifics on interventions evaluated in studies with child marriage prevention as an outcome, 2000–2019; multicomponent, single component, and multiarm
      Study, location, quality, resultsIntervention specifics
      Multicomponent
       CEDPA, 2001, India [
      • Malhotra A.
      • Amin A.
      • Nanda P.
      Catalyzing gender norm change for adolescent sexual and reproductive health: Investing in interventions for structural change.
      ], low quality; positive results
      Better Life Options Program, 1989–1999, NGO run, Evaluated 1996–1999 by NGO

      Aim of empowering girls through holistic approach

      Girls: life skills, family life education, asset-building, leadership skills, link with formal/alternative education, vocational skills.

      Community: leaders, parents mobilized through advocacy and involvement.

      Services: provider training on youth-friendly SRH
       Mathur, Mehta & Malhotra, 2004, Nepal [
      • Kalamar A.M.
      • Lee-Rife S.
      • Hindin M.J.
      Interventions to prevent child marriage among young people in low- and middle-income countries: A systematic review of the published and gray literature.
      ], low quality; mixed results
      Youth Participation Program, 2000–2003, NGO–Research Institute run; Evaluated 1999-2003 by Research Institute

      Aim of empowering youth for better ASRH through participatory approach

      Girls and boys: peer education and counseling, youth groups/clubs, safe spaces, livelihoods training

      Community: adult education, task forces, community mobilization, IEC campaigns;

      Services: provider training on youth-friendly SRH; teacher training on ASRH information and support
       Kanesnathan et al., 2008, India [
      • Chae S.
      Timing of orphanhood, early sexual debut, and early marriage in four sub-Saharan African countries.
      ], high quality; null results
      DISHA Program 2005–2006, Research Inst-NGO run; Evaluated 2005–2007 by Research Institute

      Aim of empowering youth through integrated approach to ASRH

      Girls and boys: Life skills, youth groups, and safe spaces; peer education, counseling on SRH; income generation training and credit/savings links; youth as contraceptive depot holders

      Community: community sensitization; adult groups and adult-youth groups; IEC campaigns

      Services: provider training on youth-friendly SRH services
       Shahnaz et al., 2008, Bangladesh [
      • Raj A.
      • McDougal L.
      • Silverman J.G.
      • Rusch M.L.A.
      Cross-Sectional time series analysis of associations between education and girl child marriage in Bangladesh, India, Nepal and Pakistan, 1991-2011.
      ], low quality; positive results
      ELA Program 2005–2007, NGO run; Evaluated 2005–2007 by Research Institute

      Aim to empower girls on multiple fronts—combine microfinance with skills and enabling environment

      Girls: savings and loan groups; skills training on income generation; safe spaces—life skills, socializing, reading, play; Community: Parent and community mobilization and sensitization; Services: microfinance
       UNICEF; Diop, Moreau, Benga 2008, Senegal [

      Mathur S, Greene M, Malhotra A. Too Young to Wed: Education and Action Toward Ending Child Marriage. Washington, DC: International Center for Research on Women. 2003. Available at: https://www.icrw.org/wp-content/uploads/2016/10/Too-Young-to-Wed-Education-and-Action-Toward-Ending-Child-Marriage.pdf. Accessed September 20, 2020.

      ,
      • Erulkar A.
      • Medhin G.
      • Weissman E.
      The impact and cost of child marriage prevention in rural Ethiopia.
      ], low quality; null results
      Tostan Program 1996–1999, NGO-UN run; Evaluated 2006 by Research Institute

      Aim to improve health, abandonment of FGM and child marriage

      Community: multimodule course mostly for adult women but eventually also men—on human rights and responsibilities problem solving, hygiene, health; mobilization of women and men's groups; community pledge ceremonies for not undertaking FGM, child marriage
       Erulkar and Muthengi, 2009, Ethiopia [
      • Ghimire A.
      • Samuels F.
      Continuity and change in social norms and practices around marriage and education in Nepal. London Overseas Development Institute 2014.
      ], med quality; mixed results
      Berhane Hewan Program 2004–06, Research Inst-NGO run; Evaluated 2004–2006 by Research Institute

      Empower girls through multiple ways to reduce child marriage; support married girls

      Girls: School materials for in-school and returning girls; Nonformal education, mentors, girls groups for out of school girls; Health/FP referrals; cost of card for FP services

      Family: econ incentive (goat) end of 2 years if girl not married; Community: awareness raising and consultation
       Amin et al., 2011, Bangladesh [
      • Lee-Rife S.
      • Malhotra A.
      • Warner A.
      • Glinski A.M.
      What works to prevent child marriage: A review of the evidence.
      ], low quality; null results
      Kishori Abhijan Program 2001–2003, NGO-UN run; Evaluated 2001–2003 by Research Institute

      Aim to empower adolescent girls—lower school dropout, raise economic activity, and age at marriage

      Girls: Life skills—self-esteem, leadership, gender roles/rights, health; Kishori clubs as safe spaces;

      Livelihoods training (vocational skills, teacher training);

      Community: engage parents, mobilize communities Services: microcredit
       Daniel and Nanda, 2012, India [
      • Heath R.
      • Mobarak A.M.
      Manufacturing growth and the lives of Bangladeshi women.
      ], low quality; positive results
      Prachar Program 2002–2006, NGO run; Evaluated 2008–2009 by NGO

      Aim to raise age at marriage, delay 1st birth, increase spacing to 2nd birth

      Girls and boys: Life skills 3 hours/day for 5 days—delayed marriage, spacing, SRH, STIs, spousal negotiation, decisions.

      Community: BCC for community and youth mobilization, infotainment parties, awareness raising, IEC, community leaders, parents, influential
       Stark et al., 2018, Ethiopia [
      • Hallfors D.D.
      • Cho H.
      • Rusakaniko S.
      • et al.
      The impact of school subsidies on HIV-related outcomes among adolescent female orphans.
      ], low quality; null results
      COMPASS Program 2015–2016, Academic-NGO run, Evaluated 2015–2016 by Academics

      Aim to prevent violence among adolescent girls in conflict settings

      Girls: 30 weekly life skills sessions—communication, SRH, GBV; safe spaces

      Families: 8 monthly caregiver discussion groups to improve understanding, support, attitudes
       Stark et al., 2018, DRC [
      • Handa S.
      • Peterman A.
      • Huang C.
      • et al.
      Impact of the Kenya cash transfer for orphans and vulnerable children on early pregnancy and marriage of adolescent girls.
      ], medium quality; null results
      COMPASS Program 2015–2016, Academic-NGO run, Evaluated 2015–2016 by Academics

      Aim to prevent violence among adolescent girls in conflict settings

      Girls: 32 weekly 1- to 2-hour life skills sessions from mentors, networking

      Families: 10 monthly discussion groups of parent/caregiver to improve understanding, support, attitudes.
       Bandiera et al., 2018, Uganda [
      • Dake F.
      • Natali L.
      • Angeles G.
      • et al.
      Cash transfers, early marriage, and fertility in Malawi and Zambia.
      ], med quality; positive results
      ELA Program 2008, NGO-Research Institute run; Evaluated 2010–2012 Res Institute/Academics

      Aim to empower girls by relaxing multiple human capital constraints

      Girls: ELA clubs as safe spaces; “hard” vocational skills for small-scale income generation (tailoring, poultry computing), financial literacy; “soft” life skills on SRH, menstruation, negotiation, conflict, leadership.
      Single component
       Angrist et al., 2003, Colombia [
      • Paul P.
      Effects of education and poverty on the prevalence of girl child marriage in India: A district–level analysis.
      ], medium quality; positive results
      PACES Program, 1993–1998+, Government run; Evaluated 1998–1999 by Academics

      Aim to increase secondary schooling for poor boys and girls

      CCT—Education; vouchers worth 1/2 private secondary school fees given at grade 6 and renewed through 11th grade conditional on academic performance warranting promotion to next grade
       Pande et al., 2006, India [
      • Glynn J.R.
      • Sunny B.S.
      • DeStavola B.
      • et al.
      Early school failure predicts teenage pregnancy and marriage: A large population-based cohort study in northern Malawi.
      ], medium quality; positive results
      Life Skills Program, 1998–1999, NGO-Research Inst run; Evaluated 1996–2005 by Research Institute

      Aim to keep in engaged in alternative activity to marriage; provide skills.

      Life skills: 225 hours, 1-year course—developed and implemented with community consultation; used locally drawn workers to teach with profiles mapped against Government social workers for potential scale-up
       Gulemetova, 2011, Mexico [
      • Kennedy C.E.
      • Fonner V.A.
      • Armstrong K.A.
      • et al.
      The evidence project risk of bias tool: Assessing study rigor for both randomized and non-randomized intervention studies.
      ], high quality; positive results
      Opportunidades Program (pre)2002–2004+, Government run; Evaluated 2010 by Academics

      Aim to improve human capital of families in poverty

      CCT—Education school subsidy for kids in 3–12 grades conditional on 85% attendance and grade progress; cash to moms for Grades 3–9; adolescents in Grades 10–12; higher subsidy for girls to encourage girls' education.
       Alam, Baez, Del Carpio, 2011, Pakistan [
      • Baird S.
      • McIntosh C.
      • Özler B.
      When the money runs out: Do cash transfers have sustained effects on human capital accumulation?.
      ], high quality; positive results
      Female School Stipend Program (pre)2003–2006+; Government Run; Evaluated 2011 by Academics

      Aim to improve girls middle school completion and entry to secondary education

      CCT—Education female stipends of $10 per quarter to cover costs for middle school (Grades 6–8) up to 3 years; conditional on 80% attendance
       Jensen, 2012, India [
      • Levitt-Dayal M.
      • Motihar R.
      • Kanani S.
      • Mishra A.
      Adolescent girls in India choose a better future: An impact assessment of an educational programme.
      ], high quality; positive results
      BPO Recruitment Support 2003–2006, Research Inst-NGO run; Evaluated in 2012 by Academics

      Aim to increase job opportunities to improve female early life outcomes

      Job Market Access: Recruiting services for young women to get call center jobs; three coaching sessions (5 hours) in 14 months; 3-year placement support.
       Heath and Mobarek, 2014, Bangladesh [
      • Kanesathasan A.
      • Cardinal L.J.
      • Pearson E.
      • et al.
      Catalyzing change: Improving youth sexual and reproductive health through DISHA, an integrated program in India.
      ], high quality; positive results
      Garment Factory Access 1990–2004+ Government-implemented Macro Policy; Evaluated in 2009 by Academics

      Aim to improve economic opportunities

      Job Market Access: Girls growing up in villages with access to growing number of garment factories; access assessed by commuting distant to a factory from natal village
       Hallfors et al., 2015, Zimbabwe [
      Centre for Population and Development Activities
      Adolescent girls in India choose a better future: An impact assessment.
      ], medium quality; positive results
      Structural HIV Prevention Program 2007–12, Research Institute run; Evaluated 2008–2012 by Academics

      Aim to improve HIV-related outcomes for girls through school subsidies

      CCT—Education 5 years of costs for fees, uniforms, supplies, school-based female “helper” to girls entering Grade 6 conditional on staying in school
       Handa et al., 2015, Kenya [
      • Shahnaz R.
      • Karim R.
      Providing microfinance and social space to empower adolescent girls: An evaluation of BRASC’s ELA Centres.
      ], high quality; positive results
      OVC Social Cash Transfer Program 2007–14 Government run; Evaluated 2007–2011 Academics

      Aim to provide social protection to vulnerable households with orphans

      UCT-poverty—Monthly cash 1,500–2,000 Ksh (~20% hh expenses) to ultrapoor households with at least one orphan/vulnerable child aged <18 years
       Nanda et al., 2016, India [
      • Daniel E.E.
      • Nanda R.
      The effect of reproductive health communication interventions on age at marriage and first birth in rural Bihar, India: a retrospective study. Watertown: Pathfinder International.
      ], medium quality; null results
      ABAD (Our Daughter Our Wealth) Program 1994–1998 Government Run; Evaluated 2015 by Research Inst

      Aim to improve value of girls, prevent sex selection, delay marriage

      Conditional Asset transfer—Rs 500 cash at birth of girl plus Rs 25,000 bond in her name cashable at age 18 years—provided she stays unmarried. Effect assessed 18 years later on female cohorts that benefited at birth.
       Dake et al., 2018, Malawi & Zambia [
      • Glick P.
      • Handy C.
      • Sahn D.E.
      Schooling, marriage, and age at first birth in Madagascar.
      ], high quality; null results
      Social Cash Transfer Programs 2011–2017 Zambia; 2006–2020 Malawi Govt run

      Evaluated Zambia: 2011–2013; Malawi: 2013–2015 by Research Institute

      Aim to support ultrapoor, labor-constrained households on basic needs

      UCT-poverty—cash to households equal to approximately 20% of consumption. Malawi: bimonthly transfer: 2,000–4,800 Kwacha (US $5.80–$13.30). Zambia—bimonthly transfer of 120 kwacha (US $24); effect assessed on girls 14–21 years.
       Koski et al., 2018, 8 African countries [
      • Gulemetova M.
      Evaluating the impact of conditional cash transfer programs on adolescent decisions about marriage and fertility: The case of Oportunidades.
      ], medium quality; mixed results
      Primary school fee elimination 1995–2002 Govt macro policy; Evaluated 2016–2017 by Academics

      Aim to increase primary school enrollment

      School fee elimination 1995–2002 at primary level in eight African countries vis a vis eight control countries, which acted later. Effect assessed on cohorts that experienced fee reduction
       Hahn et al., 2018, Bangladesh [
      • Nanda P.
      • Datta N.
      • Das P.
      • et al.
      Making change with cash? Impact of a conditional cash transfer program on age of marriage in India.
      ], high quality; positive results
      Female Secondary School Stipend Program 1994–2001+ Government run; Evaluated 2015–2016 by Academics

      Aim to improve 6–10 grade completion, labor market options, delay marriage

      CCT-education—annual stipend of $18 for Grade 6 to $45 for Grade 10 conditional on 75% attendance, passing test scores, staying unmarried. Fees paid directly to school; amount for other costs in girl's bank account.

      Multiarm
       Baird, McIntosh & Ozler, 2009, 2011, and 2016, Malawi [
      • Erulkar A.S.
      • Muthengi E.
      Evaluation of Berhane Hewan: A program to delay child marriage in rural Ethiopia.
      ,
      • Diop N.J.
      • Moreau A.
      • Benga H.
      Evaluation of the long-term impact of the TOSTAN programme on the abandonment of FGM/C and early marriage: Results from a qualitative study in Senegal.
      ,
      UNICEF
      Long-term evaluation of the Tostan programme in Senegal: Kolda, Thiès and Fatick regions.
      ], high quality; mixed results
      Zomba Cash Transfer Program 2007–2009 Academic--NGO run; Evaluated 2007–2016 by Academics

      Aim to improve human capital accumulation, related life experiences especially for girls

      Arm 1: CCT-education—school fees and $10 subsidy 10 months per year for 2 years conditional on 80% school attendance to girls who were school dropouts at baseline (age 13–22 years)

      Arm 2: UCT-equivalent cash, no condition—girls aged 13–22 years in school at baseline

       Duflo, Dupas, & Kremer, 2015, Kenya [
      • Erulkar A.
      • Medhin G.
      • Weissman E.
      The impact and cost of child marriage prevention in rural Burkina Faso.
      ], high quality; mixed results
      School Subsidy & HIV Education Program 2003–2004 Academic-Govt run; Evaluated 2003–2010 by Academics

      Aim to test policy instruments to reduce early pregnancy and STI risk

      Arm 1. CCT—Education uniforms for up to 2 years (worth $12) for Grades 6-7 conditional on staying in school

      Arm 2. HIV EducationAbstinence-only teacher training for government programs in schools

      Arm 3. Multicomponent (Arms 1 and 2)
       Amin et al., 2016; 2018, Bangladesh [
      • Heath R.
      • Mobarak A.M.
      Manufacturing growth and the lives of Bangladeshi women.
      ,
      • Stark L.
      • Seff I.
      • Asghar K.
      • et al.
      Building caregivers’ emotional, parental and social support skills to prevent violence against adolescent girls: Findings from a cluster randomised controlled trial in Democratic Republic of Congo.
      ], high quality; positive results
      Balika Program 2014–2015 Research Institute-NGO run; Evaluated 2014–2015 by Research Institute

      Aim to delay age at marriage and empower girls

      Arm 1: Education support—tutoring for in-school girls; computing or financial training for out-of-school girls

      Arm 2: Gender rights awareness training

      Arm 3: Livelihoods training in computers, servicing, first aid, photography
       Erulkar Medhin and Weissman, 2017, Ethiopia [
      • Bandiera O.
      • Buehren N.
      • Burgess R.
      • et al.
      Women’s empowerment in action: Evidence from a randomized control trial in Africa.
      ], low quality; mixed results
      Berhane Hewan Phase 2 2013–2014 Research Institute-NGO-Government run; Evaluated 2013–2014 Research Institute

      Aim to prevent child marriage and empower girls

      Arm 1: Community dialog—weekly discussion groups-harms of child marriage

      Arm 2: CCT-Education school supplies condition staying in school, not marry

      Arm 3: Conditional Asset Transfer—two chickens per year conditional on not marrying and staying in school

      Arm 4: Multicomponent—all three arms combined

      No control site as it was compromised
       Erulkar et al., 2017, Tanzania [
      • Angrist J.
      • Bettinger E.
      • Bloom E.
      • et al.
      Vouchers for private schooling in Colombia: Evidence from a randomized natural experiment.
      ], medium quality; mixed results
      Berhane Hewan Replication 2013–2014 Research Institute-NGO-Govt run; Evaluated 2013–2014 by Research Institute

      Aim to prevent child marriage and empower girls

      Arm 1: Community sensitization—discussions and info on harms of marriage, at village meetings, events

      Arm 2: CCT-Education—school supplies condition staying in school, not marry

      Arm 3: Conditional Asset Transfer—livestock at the end of intervention conditional on not marrying and staying in school

      Arm 4: Multicomponent—all three arms combined
       Erulkar et al., 2017

      Burkina Faso [
      • Pande R.
      • Kurz K.
      • Walia S.
      • et al.
      Improving the reproductive health of married and unmarried youth in India.
      ], med quality; mixed results
      Berhane Hewan Replication 2013–2014 Research Institute-NGO-Govt run; Evaluated 2013–2014 by Research Institute

      Aim to prevent child marriage

      Aim to prevent child marriage and empower girls

      Arm 1: Community dialog—weekly discussion groups-harms of child marriage

      Arm 2: CCT-Education school supplies condition staying in school, not marry

      Arm 2: Multicomponent--2 arms combined
       Buchmann et al., 2018, Bangladesh [
      • Jensen R.
      Do labor market opportunities affect young women’s work and family decisions? Experimental evidence from India.
      ], high quality; mixed results
      Kishori Kontha (pre) 2008–2010 Academic-NGO run; Evaluated 1995–2002 by Academics

      Aim to reduce child marriage, teen childbearing, increase education<

      Arm 1: Rights-based life skills training—6 months, 5–6 days 2 hours/day—life skills, literacy, numeracy, SRH, communication, safe spaces, social activities

      Arm 2: Conditional Asset transfer: cooking oil worth $16 per year, conditional on girls not marrying till 18 years

      Arm 3: Multicomponent—Arms 1 and 2
      Table 1 shows that most of the multicomponent (8/11) and single-component (8/12) studies used a quasi-experimental design with cluster randomized controlled trials (RCTs) a less frequent occurrence. Four of the seven multiarm studies, on the other hand, were RCTs, with the remaining three using a quasi-experimental approach. Multicomponent studies were much more likely to be low quality (7/11), whereas there were no low-quality single-component studies and only one low-quality multiarm study.
      Table 1 also outlines the different outcome measures studies use to assess the impact on early or child marriage: (1) incidence of marriage before age 18 years, (2) incidence of ever marriage among girls and young women, and (3) mean/median age of marriage. To assess program effects on the incidence of marriage before age 18 years or ever marriage, studies used several analytic measures, most frequently odds ratios, risk ratios, and hazard rates/ratios or compared proportions married across treatment and nontreatment groups. Others examined program effects through comparing mean/median age at marriage, examining whether the intervention succeeded in delaying entry into marriage. This is important to note because although some programs had no effect on reducing child marriage incidence, they were successful in delaying age of entry into marriage.
      As studies often do not provide information on the percentage of girls at risk of child marriage in the contexts where the intervention was implemented, we used a reach of 25,000 participants as a conservative threshold for scale. Table 1 shows that only 2 of the 11 multicomponent interventions met this threshold, and only one of the seven multiarm programs did so. In contrast, 9 of the 12 single-component interventions reached 25,000 or more participants, with eight of these nine programs reaching several hundred thousand or even millions of participants. Similarly, the single-component programs show a better sustainability record (8/12) than the multicomponent (2/11) and multiarm (1/7) programs. Table 2 clarifies the reason for this pattern: although all the multicomponent and multiarm programs were designed and implemented by NGOs and/or research organizations, 9 of the 12 single-component programs or policies were implemented by governments. In fact, the only two multicomponent programs that reached more than 25,000 participants were exceptional in being implemented by BRAC, the world's largest NGO.
      Table 2 shows that multicomponent interventions covered in our analysis go back to the early 1990s. A multipath approach to empowering girls was the focus in seven programs. Six of these programs include life skills, safe spaces or adolescent clubs, and vocational training or livelihood skills for girls as core components. Five programs—all in Asia—were very comprehensive in also including multiple family and community mobilization activities along with a service component of either teacher or health provider training or access to microcredit. The Berhane Hewan program in Ethiopia used a slightly different multipath combination of formal and nonformal schooling support to girls, asset transfer of livestock conditional on girls not marrying, health service vouchers, and parental and community mobilization. Four other multicomponent programs had narrower goals and used fewer components, from targeting the abandonment of female genital mutilation and child marriage through community education and village pledge events in Senegal, to combining life skills and parental engagement to reduce gender-based violence among girls in humanitarian settings in Africa. The Prachar program in India used limited sexual and reproductive health education sessions in combination with behavioral change communication activities with adolescents and communities to delay marriage and childbearing.
      As indicated by Table 2, single-component programs in our review also go back to the early 1990s, with the most common intervention (5/12) being some type of conditional cash or in-kind transfer for school attendance. Four of these programs were implemented by governments, two each in Latin America and Asia, whereas the one in Africa was a Research-NGO partnership. The main motivation of the government programs was to improve the school success and human capital of children living in poverty—especially girls—with delayed marriage as a potential secondary outcome. The smaller scale conditional cash transfer (CCT) for school program in Zimbabwe aimed to reduce HIV risk for orphaned girls through school retention and delayed sexual debut. The programs in Asia and Africa focused specifically on girls to address gender disparities, whereas those in Latin America covered both boys and girls. In Mexico, all schooling levels were included, but subsidies were higher for girls in secondary school. The Colombia, Bangladesh, and Zimbabwe interventions were for secondary school, whereas the Pakistan intervention was for primary school.
      Three of the other single-component studies also used cash or asset transfers but differently. Unconditional cash transfer (UCT) programs in Kenya, Malawi, and Zambia were government efforts at mitigating poverty and providing a safety net of approximately 20% monthly income to ultrapoor families in the context of HIV/AIDS in Africa. These programs were not intended to delay marriage, but evaluators considered improved welfare a potential path for positive adolescent outcomes. An Indian state government's long-term conditional asset transfer program, on the other hand, aimed to prevent both male-favored sex selection and child marriage by depositing a bond at a girl's birth and making it cashable 18 years later if she remained unmarried.
      Two single component studies examined the impact of macro policies increasing access for young women to call center jobs in India and garment sector jobs in Bangladesh. The increased economic opportunity resulting from government policy action could be expected to delay marriage by shifting girls' and parental aspirations toward an attractive alternative to early marriage. Both studies considered the proximity of girls' and young women's homes to call centers and factories, but the India study also provided job placement support. Another study examined government policy action on eliminating primary school fees in eight African countries, with the main aim of not delaying marriage but facilitating basic education for all children. However, evaluators explored the policy action as a potential catalyst for preventing child marriage.
      Among the seven multiarm studies, the most common intervention arm (5/7) was also a conditional cash or in-kind subsidy for girls to attend school. In the context of HIV/AIDS and improving adolescent life outcomes, one study in Malawi compared this arm with an arm on UCTs, whereas another in Kenya compared it with abstinence-only training for teachers and also tested the impact of the two components combined. Three of the multiarm studies in Africa tried to unpack the Berhane Hewan multicomponent approach previously tried in Ethiopia by testing separately the impact of conditional schooling support to girls in Ethiopia, Burkina Faso, and Tanzania. Two of these studies--in Ethopia and Tanzania--also tested asset transfers conditional on a girl not marrying. Although these programs in Africa used livestock as the asset, the Kishori Kontha program in Bangladesh conditioned the transfer of cooking oil as an asset conditional on girls not marrying.
      The two multiarm studies in Bangladesh tested the stand-alone impact of some type of a gender rights or rights-focused life skills training curriculum on preventing child marriage. One of these studies also tested this intervention in combination with asset transfers as a multicomponent arm. The evaluation of the Balika program only tested stand-alone interventions; in addition to a gender rights training arm, it also included arms on livelihood skills and on education support to girls through tutoring, computer, or financial skills. The three Berhane Hewan replication programs in Tanzania, Ethiopia, and Burkina Faso each tested community mobilization as a separate arm and also had an arm with all components bundled together.

      Impact on child marriage

      In Table 3A, Table 3B, Table 3C, we present findings from the 30 studies with regard to impact on child and early marriage prevention, separately for multicomponent, single component, and multiarm interventions, and by subcategories within these groupings.
      Table 3AResults from 11 multicomponent intervention evaluations on preventing child marriage: by intervention focus
      Multicomponent studyCountryResultSummary
      Girl's empowerment focus
       CEDPA 2001IndiaPositivePositive: 3 of 7

      Positive medium- to high-quality studies:1 of 3
       Mathur et al., 2004IndiaMixed
       Kanesthasan et al., 2006 (Q)IndiaNull
       Shahnaz & Karim, 2008 (S)BangladeshPositive
       Erulkar & Muthengi, 2009 (Q)EthiopiaMixed
       Amin et al., 2011BangladeshNull
       Bandiera et al., 2018 (Q) (S)UgandaPositive
      Other programmatic focus
       UNICEF, 2008SenegalNullPositive: 1 of 4

      Positive medium- to high-quality studies:0 of 1
       Daniel and Nanda, 2012IndiaPositive
       Stark et al., 2018aEthiopiaNull
       Stark et al., 2018b (Q)DRCNull
      Positive results all studies: 4 of 11; Positive results all medium- and high-quality studies: 1 of 4
      Q = medium- or high-quality studies; S = scale criterion of interventions with 25,000 or more participants.
      Table 3BResults from 12 single component intervention evaluations on preventing child marriage: by intervention type
      Single Component StudyCountryTypes of cash or asset transfersOther interventions
      CCT school supportUCT poverty alleviationAsset transfer condition not marryJob market accessRights/life skillsSchool fee elimination
      Angrist et al., 2003 (Q) (S)ColombiaPositive
      Pande et al., 2006 (Q)IndiaPositive
      Gulemetova 2011, (Q) (S)MexicoPositive
      Alam et al. 2011, (Q) (S)PakistanPositive
      Jensen, 2012 (Q)IndiaPositive
      Heath & Mobarek, 2014 (Q) (S)BangladeshPositive
      Hallfors et al., 2015 (Q)ZimbabwePositive
      Handa et al., 2015 (Q) (S)KenyaNull
      Nanda et al., 2016 (Q) (S)IndiaNull
      Dake et al., 2018 (Q) (S)Malawi, ZambiaNull
      Koski et al., 2018 (Q) (S)8 African countriesMixed
      Hahn et al., 2018 (Q) (S)BangladeshPositive
      Results 12 Studies: 8 Positive: 1 Mixed; 3 Null

      (All Medium-High Quality)
      Positive Results by Intervention5 of 50 of 20 of 12 of 21 of 10 of 1
      CCT = conditional cash transfer; Q = medium- or high-quality studies; S = scale criterion of interventions with 25,000 or more participants; UCT = unconditional cash transfer.
      Table 3CResults from seven multiarm intervention evaluations, by intervention arm
      Multiarm StudyCountryMulticomponentTypes of cash or asset transfersOther Interventions
      Multipath girls' empowermentOther approachesCCT school supportUCT poverty AlleviationAsset transfer condition not marryingLivelihood skillsGender rights training; Rights/life skillsCommunity mobilizationAbstinence-only teacher training
      Duflo et al., 2015 (Q)KenyaNegativePositiveNull
      Baird et al., 2016 (Q)MalawiPositiveMixed
      Amin et al., 2016 (Q)BangladeshPositivePositivePositive
      Erulkar et al., 2016aEthiopiaMixedMixedMixedMixed
      Erulkar et al., 2016b (Q)TanzaniaMixedNullPositiveNull
      Erulkar et al., 2016c (Q)Burkina FasoNullNullNegativePositive
      Buchmann et al., 2018 (Q) (S)BangladeshNullPositiveNull
      Positive results per arm all seven studies0 of 40 of 13 of 60 of 12 of 41 of 11 of 21 of 30 of 1
      Positive results per arm medium- to high-quality studies0 of 30 of 13 of 50 of 12 of 31 of 11 of 21 of 20 of 1
      CCT = conditional cash transfer; Q = medium- or high-quality studies; S = scale criterion of interventions with 25,000 or more participants; UCT = unconditional cash transfer.
      Because the content and theory of change for multicomponent programs using a multipath approach to empowering girls is likely to differ from multicomponent programs using other approaches, in Table 3A, we examine the findings for the seven girls' empowerment-focused multicomponent studies separately from the four studies on bundled interventions with other approaches. Only three of the seven empowerment-focused multicomponent interventions show positive findings; two studies show mixed results, and two studies show no effect. When we eliminate the four low-quality studies in this group, we are left with only one out of three studies with positive findings. This intervention—the ELA program in Uganda—also meets our threshold for scale in reaching more than 25,000 participants. Of the four multicomponent studies with other approaches, only one—the Prachar program in India—has positive results. But as this study is low quality, there are no studies with a positive effect in this subgroup once the sensitivity analysis is applied.
      In Table 3B, we turn to single-component interventions, and keeping in mind the important differences in the structure, content, and pathways to change among the different types of cash transfer programs, we first examine the results for conditional cash or in-kind transfer programs supporting girls' schooling. These aim to provide an alternative to marriage and build long-term human capital. All five of the programs in this subcategory show positive results. Moreover, four of these programs—two in Latin America and two in Asia—were government-run interventions that more than meet our criteria for scale. In contrast, both the studies of government-run UCT programs in Africa show no effect. The one long-term asset transfer program conditional on delaying marriage till 18 years and run by a state government in India, also shows no effect.
      Looking next at the two studies assessing the impact of favorable job markets presented by call centers in India and the garment industry in Bangladesh, both show positive effects and also meet the scale criteria. The study in India on the stand-alone impact of life skills in delaying child marriage also has positive findings but does not meet the scale threshold. (However, the life skills program was eventually scaled up in an adapted version with the Maharashtra government.) The one study on school fee elimination across multiple countries in Africa was at scale and sustained but shows mixed effects. As there were no low-quality studies among the single-component evaluations, the findings in Table 3B are robust against our sensitivity analysis.
      In Table 3C, we review the results from the seven multiarm studies. We first consider the five studies with multicomponent arms, four of which took a multipath approach to girls' empowerment. None of these arms showed positive results, with two studies in Ethiopia and Tanzania showing mixed results. In the fifth study in Kenya which took a different approach, the multicomponent intervention arm actually increased early marriage. Moreover, in all four medium-high quality multiarm studies that tested a combination of single and multicomponent arms, the single-component arms performed better than the multicomponent arms.
      Conditional cash or in-kind support for girls' schooling was the most common intervention tested across six of the seven multiarm studies: three studies had positive results, one mixed results, and two showed no impact. With sensitivity analysis, the ratio improves to three of five studies with positive results for this intervention arm. Four studies included arms on the impact of asset transfers conditional on girls not marrying, with a diversity of results: two positive, one mixed, and one negative where child marriage increased. Sensitivity analysis also improves these results, with two of three positive findings, one for the Berhane Hewan replication in Tanzania, and one for the Kishori Kontha program in Bangladesh. The latter is also the only successful intervention to meet our scale criteria within all multiarm studies.
      The Kishori Kontha evaluation in Bangladesh found no effect for its stand-alone rights-focused life skills arm, but the Balika evaluation found a positive effect for both its gender rights training and livelihoods training arms. The stand-alone community mobilization interventions in the three Berhane Hewan program replications in Africa found no consistent results: one positive, one null, and one mixed. The Kenya study with an arm on abstinence-only training for teachers found no impact.
      Table 4 consolidates findings on success rates from evaluation arms in multiarm studies with evaluation findings for similar interventions in the multicomponent and single-component studies. As such, it provides a comprehensive summary of success rates for different subcategories of interventions across all studies, both with and without sensitivity analysis. The table consolidates results from 11 evaluations of multipath girls' empowerment-focused multicomponent interventions—a number large enough to assess a clear pattern on findings. The overall success rate for these interventions is only 3 in 11, and because almost half the evaluations in this category are low quality, sensitivity analysis leaves us with a success rate of only one in six. Multicomponent interventions with other approaches perform even more poorly, with an overall success rate of only one in 5, and with sensitivity analysis further down to zero positive results from two studies.
      Table 4Consolidated success rate of evaluations to prevent child and early marriage from multicomponent, single component, and multiarm studies by specific intervention type and evaluation quality
      Grouped intervention categoriesSpecific intervention typeShare of positive findings in all studies in categoryShare of positive findings in medium- to high-quality studies
      Multicomponent1. Multipath approach to girls' empowerment3 of 111 of 6
      2. Other bundled approaches1 of 50 of 2
      Single component3. Conditional cash or kind transfer for schooling support8 of 118 of 10
       Cash or asset4. Asset transfer conditional on delayed marriage2 of 52 of 4
       Transfer5. Unconditional cash transfer for poverty alleviation0 of 30 of 3
      Single component6. Female job opportunities or livelihood skills training3 of 33 of 3
       Economic opportunity or skills7. Gender rights/life skills training2 of 32 of 3
      Single component8. Community mobilization1 of 31 of 2
       Other9. School fee elimination0 of 10 of 1
       Interventions10. Abstinence-only education0 of 10 of 1
      Among single-component interventions, the impact of CCTs for education on child marriage prevention is also assessed in 11 evaluations and demonstrates a much higher success rate of 8 in 11, improving to 8 in 10 with the elimination of one low-quality evaluation. Although most other single-component interventions have too few evaluations to provide the critical mass for determining a clear pattern, female labor market opportunities and livelihood skills show some promise with all three high-quality studies showing a positive impact. Similarly, with a success rate of 2 in 3, there may be promise in gender rights and life skills trainings as stand-alone interventions. In contrast, 0 of 3 positive findings for UCT programs suggests that this approach may be less promising in addressing child marriage.

      Discussion

      In summary, with a critical mass of 30 evaluations over the last 20 years, we are able to reach clearer conclusions on what works to delay marriage than it has been possible in the past. With 16 studies not covered by past reviews (13 published since 2016), our analysis covers a broader range of interventions, evaluation methodologies, and a higher share of more rigorous evaluations, allowing for more substantiated findings. Especially helpful is the increased share of single-component and multiarm studies, which allow for greater specificity in comparing program or policy strategies, as well as a higher share of large-scale, government-delivered programs and policies, allowing for a more deliberate analysis of the reach and sustainability of different approaches.
      We see the most definitive pattern of success in preventing child marriage among interventions that supported girls' school attendance and progress through cash or in-kind transfers, with 8 out of 11 studies showing positive results, and sensitivity analysis increasing this ratio to 8 out of 10. Because of the larger cumulative pool of evaluations, we were able to differentiate the results for CCTs on school support from the impact of other “economic incentive” interventions most often conflated together in the literature on child marriage prevention. Comparatively, we find that asset transfer interventions with delayed marriage as a condition show only a 50% rate of success. Given the variations in context, design, and rigor of the five studies evaluating this approach, further understanding and evidence is needed for more definitive results on when an asset transfer might work and when it might not. For example, in the India study where the girls and families had to wait 18 years before receiving the transfer, part of the reason for the lack of effect was that secular declines in child marriage across the state had already overridden any effect the program could have. In the Berhane Hewan replication in Africa, the asset transfer of livestock was a strong enough incentive in some settings, but not in others, whereas cooking oil in Bangladesh seemed to be a good enough incentive for girls and their families to keep them single.
      We also find that the three evaluations of UCTs intended for poverty mitigation—all in SSA—show null results. In many ways, this is not surprising because the programs were intended to alleviate extreme poverty rather than delay marriage. It is possible that the cash households received was spent on basic needs or other priorities rather than providing sufficient advantage to girls for their marriage to become undesirable. In their multiround study comparing CCTs for school support with UCTs for household economic support, Baird et al. posit that with UCTs, “when the cash runs out,” there has been no value added for girls, whereas with the CCT, staying in school is forging a different life trajectory for them [
      • Daniel E.E.
      • Nanda R.
      The effect of reproductive health communication interventions on age at marriage and first birth in rural Bihar, India: a retrospective study. Watertown: Pathfinder International.
      ]. Thus clearly, all cash or asset transfers cannot be considered together in assessing their impact on child marriage.
      These results suggest that supporting girls' schooling through cash or in-kind transfers is proving to be in fact the most successful channel for delaying their marriage among the programs evaluated to date and by far more effective than economically supporting their families or motivating them through cash or in-kind incentives. That enhancement of girls' own human capital is a compelling pathway to delaying marriage is also supported by a high share of positive results among the few studies that assessed an exclusive focus on life skills, livelihoods, and gender rights training for girls.
      In addition, the very large positive effects in two studies on favorable job markets for women in India and Bangladesh provide some indication that not just investment in, but the visible promise of economic opportunities for girls in early adulthood, is important in delaying marriage. Heath and Mobarek note, for example, that by their estimates, the impact of the presence of garment factories in Bangladesh is many times larger in delaying marriage than even the massive government-supported female secondary school stipend program [
      • Stark L.
      • Asghar K.
      • Seff I.
      • et al.
      Preventing violence against refugee adolescent girls: Findings from a cluster randomised controlled trial in Ethiopia.
      ].
      Besides their effectiveness, targeted interventions that enhance girls' human capital and their employment opportunities should also be attractive for advocates and practitioners because a more significant share of them operate at scale. For example, four of the eight programs with positive results on CCTs for school support were large-scale government implemented efforts, reaching many thousands and millions of girls. Similarly, macro policies supporting female employment in the garment industry in Bangladesh or the outsourcing boom in India also affected girls and young women in very large numbers. From the available evidence, these two intervention approaches demonstrate a combination of success as well as scale and sustainability more extensively than any other category of interventions.
      With a critical mass of studies now available on multicomponent interventions, our analysis also highlights the low success rate of comprehensive programs in preventing child marriage. Our findings show that most multicomponent programs that try to empower girls through multiple pathways are not successful in preventing child marriage, with only 3 of 11 programs demonstrating positive results, and only one out of six higher quality evaluations doing so. Multicomponent programs that do not take the empowerment approach are even less successful. As these programs are rarely at scale or sustained, these findings raise important questions regarding continued advocacy for and investment in this approach, especially for those in the global community who are struggling to accelerate action toward achieving the SDG target 5.3.1 within the next 10 years.
      Although limited in number, the results from multiarm studies are particularly interesting in this regard as they often show the single components of bundled programs to be more effective than the combination. It is possible that the stand-alone interventions are more intense, or of longer duration, or implemented at higher quality. A few studies mention the challenge of setting up and fully delivering integrated interventions within the typical 2- to 3-year period available to most programs [
      • Mathur S.
      • Mehta M.
      • Malhotra A.
      Youth reproductive health in Nepal: Is participation the answer?.
      ,
      • Stark L.
      • Seff I.
      • Asghar K.
      • et al.
      Building caregivers’ emotional, parental and social support skills to prevent violence against adolescent girls: Findings from a cluster randomised controlled trial in Democratic Republic of Congo.
      ]. There are also some hints in studies that the uptake of stand-alone interventions may be higher, given the demands of multifaceted programs on girls' and their families' time [
      • Bandiera O.
      • Buehren N.
      • Burgess R.
      • et al.
      Women’s empowerment in action: Evidence from a randomized control trial in Africa.
      ]. More systematic documentation and analysis of implementation processes are needed to better understand why multicomponent programs are not succeeding at higher rates. Equally, a larger pool of multiarm studies might be more conclusive in establishing that single—and perhaps the strongest—interventions from the comprehensive programs might be the better option.
      In considering school-focused CCTs and possibly female-centered employment policies as the logical priority based on our findings, it will be important to explore how child marriage prevention might be integrated as a central rather than peripheral focus in such interventions and policies. For example, in working with the education sector, factors such as targeting of primary versus secondary school, private versus public schools, just girls or both girls and boys, would be important considerations, as successful interventions to date span all these parameters. Moreover, the success of demand-side interventions to promote girls' schooling through cash and in-kind support begs the question of the potential positive effects of supply-side interventions for girls' schooling, such as more secondary schools or female teachers, better curricula and skills, or more accessible transportation. While education sectors in many countries are certainly implementing such initiatives, currently, no evaluations examine the impact on child marriage of these supply-side interventions to promote girls' schooling outcomes, a gap that would be important for researchers and advocates to address.
      Our review should be considered in light of some limitations. Although broad, our search strategy may have missed some evaluations, especially those published in languages other than English. Second, despite tapping all available sources, we were constrained both by space considerations and by inconsistency in details available for each study to integrate all the parameters that would be desirable for an even more comprehensive analysis. Several studies did not provide sufficient information regarding implementation science questions, which warrant consideration, including information on intervention dosage and intensity, fidelity, unintended consequences, and, importantly, cost and cost-effectiveness. Only a few studies reported unintended or negative results, which could not be assessed systematically.
      In addition, the variability of outcome measures and age groups of participants made it difficult to make quantitative comparisons across studies, limiting conclusions about the magnitude of the impact of child marriage interventions. Several studies examined probability or proportion of marriage among girls under age 18 years who have not completed time at risk, and thus, there is a strong possibility that child marriage prevalence was underestimated because of right censoring. Most included evaluations relied on retrospective self-reporting of age at marriage, and there is a possibility of measurement error, including recall and social desirability bias, especially in places where child marriage is illegal.
      Space considerations constrain us from an analysis of other behavioral outcomes closely related to child marriage that several studies included—such as school retention or pregnancy reduction as well as results on nonbehavioral outcomes such as knowledge and attitudes, which may be related to child marriage results indirectly and over the long term.

      Acknowledgments

      An oral presentation on progress on this research was made at an online webinar on “Advancing the evidence base on strategies to end child marriage and support married girls: outcomes of the 2019 Geneva meeting and next steps,” hosted by the World Health Organization and Girls Not Brides. The authors are thankful for the comments and feedback received from webinar participants.

      Funding Sources

      This research received partial financial support from the UNFPA-UNICEF Global Programme to End Child Marriage.

      References

        • Santhya K.G.
        Early marriage and sexual and reproductive health vulnerabilities of young women: A synthesis of recent evidence from developing countries.
        Curr Opin Obstet Gynecol. 2011; 23: 334-339
        • Marphatia A.A.
        • Ambale G.S.
        • Reid A.M.
        Women’s marriage age matters for public health: A review of the broader health and social implications in South Asia.
        Front Public Health. 2017; 5: 269
        • Koski A.
        • Clark S.
        • Nandi A.
        Has child marriage declined in Sub-Saharan Africa? An analysis of trends in 31 countries.
        Popul Dev Rev. 2017; 43: 7-29
        • Olamijuwon E.O.
        • Chisumpa V.H.
        • Akinyemi J.O.
        Unveiling the realities of marrying too young: Implications of child marriage on sexual and reproductive health of girls and infant survival in Sub-Sahara Africa.
        Afr Popul Stud. 2017; 31 ([Internet]): 3594-3610
        • Godha D.
        • Gage A.J.
        • Hotchkiss D.R.
        • Cappa C.
        Predicting maternal health care use by age at marriage in multiple countries.
        J Adolesc Health. 2016; 58: 504-511
        • Delprato M.
        • Akyeampong K.
        The effect of early marriage timing on women’s and children’s health in Sub-Saharan Africa and Southwest Asia.
        Ann Glob Health. 2017; 83: 557-567
        • UNICEF
        Child marriage [Internet].
        (Available at:)
        • Shapiro D.
        • Gebreselassie T.
        Marriage in Sub-Saharan Africa: Trends, determinants, and consequences.
        Popul Res Policy Rev. 2014; 33: 229-255
        • Irani M.
        • Roudsari R.L.
        Reproductive and sexual health consequences of child marriage: A review of literature.
        J Midwifery Reprod Health. 2019; 7: 1584-1590
        • Raj A.
        When the mother is a child: The impact of child marriage on the health and human rights of girls.
        Arch Dis Child. 2010; 95: 931-935
        • Malhotra A.
        • Elnakib S.
        Evolution in the evidence base on child marriage 2000-2019.
        (Forthcoming)2020
        • Girls not brides
        SDGs and Child Marriage.
        2019
        • UNICEF
        Child marriage: Latest trends and future prospects.
        2018
        • UNICEF
        Ending child marriage: Progress and prospects.
        ([Internet]. UNICEF DATA. 2014 Available at:)
        • UNICEF
        Child marriage [Internet].
        (UNICEF DATA. Available at:)
        • United Nations
        Goal 5: Sustainable development knowledge platform [Internet]. Sustainable development goals.
        (Available at:)
        https://sustainabledevelopment.un.org/sdg5
        Date accessed: June 14, 2020
        • Finnie A.
        • Malhotra A.
        • Travers E.
        • et al.
        Advancing the evidence base on strategies to end child marriage & support married girls. Conference Proceedings. 2019 Sept 30-Oct 2; Geneva, Switzerland.
        (Available at:)
        • Svanemyr J.
        • Chandra-Mouli V.
        • Raj A.
        • et al.
        Research priorities on ending child marriage and supporting married girls.
        Reprod Health. 2015; 12 ([Internet]): 80
        • Forte C.L.
        • Plesons M.
        • Branson M.
        • Chandra-Mouli V.
        What can the global movement to end child marriage learn from the implementation of other multi-sectoral initiatives?.
        BMJ Glob Health. 2019; 4: e001739
      1. Lessons learned from national initiatives to end child marriage - 2016 [Internet]. Girls Not Brides.
        (Available from:)
        • Jain S.
        • Kurz K.
        New insights on preventing child marriage: A global analysis of factors and programs.
        International Center for Research on Women (ICRW), New Delhi, India2007
        • Das Gupta S.
        • Pande R.
        Knot ready: Documenting initiatives to delay early marriage: Update 1.
        Int Cent Res Women ICRW, Washington DC2008
        • Kalamar A.M.
        • Lee-Rife S.
        • Hindin M.J.
        Interventions to prevent child marriage among young people in low- and middle-income countries: A systematic review of the published and gray literature.
        J Adolesc Health. 2016; 59: S16-S21
        • Lee-Rife S.
        • Malhotra A.
        • Warner A.
        • Glinski A.M.
        What works to prevent child marriage: A review of the evidence.
        Stud Fam Plann. 2012; 43: 287-303
        • Chae S.
        • Ngo T.
        The global state of evidence on interventions to prevent child marriage.
        2017
        • Ghimire A.
        • Samuels F.
        Continuity and change in social norms and practices around marriage and education in Nepal. London Overseas Development Institute 2014.
        (Available at:)
        • Steinhaus M.
        • Hinson L.
        • Rizzo A.T.
        • Gregowski A.
        Measuring social norms related to child marriage among adult decision-makers of young girls in Phalombe and Thyolo, Malawi.
        J Adolesc Health. 2019; 64: S37-S44
        • Cislaghi B.
        • Mackie G.
        • Nkwi P.
        • Shakya H.
        Social norms and child marriage in Cameroon: An application of the theory of normative spectrum.
        Glob Public Health. 2019; 14 ([Internet]) ((Nkwi P.) Catholic University of Cameroon, Bamenda, Cameroon): 1479-1494
        • Kenny L.
        • Koshin H.
        • Sulaiman M.
        • Cislaghi B.
        Adolescent-led marriage in Somaliland and Puntland: A surprising interaction of agency and social norms.
        J Adolesc. 2019; 72: 101-111
        • Mathur S.
        • Mehta M.
        • Malhotra A.
        Youth reproductive health in Nepal: Is participation the answer?.
        International Center for Research on Women, New York2004
        • Amin S.
        Empowering adolescent girls in rural Bangladesh: Kishori Abhijan.
        2011
        • Levitt-Dayal M.
        • Motihar R.
        • Kanani S.
        • Mishra A.
        Adolescent girls in India choose a better future: An impact assessment of an educational programme.
        2003
        • Erulkar A.S.
        • Muthengi E.
        Evaluation of Berhane Hewan: A program to delay child marriage in rural Ethiopia.
        Int Perspect Sex Reprod Health. 2009; 35: 6-14
        • Glick P.
        • Handy C.
        • Sahn D.E.
        Schooling, marriage, and age at first birth in Madagascar.
        Popul Stud. 2015; 69: 219-236
        • Adu Boahen E.
        • Yamauchi C.
        The effect of female education on adolescent fertility and early marriage: Evidence from free compulsory universal basic education in Ghana.
        J Afr Econ. 2018; 27: 227-248
        • Rasmussen B.
        • Maharaj N.
        • Sheehan P.
        • Friedman H.S.
        Evaluating the employment benefits of education and targeted interventions to reduce child marriage.
        J Adolesc Health. 2019; 65: S16-S24
        • Wodon Q.T.
        • Jesse C.
        • Belhaj N.
        • et al.
        Tanzania economic update: The power of investing in girls-Educating girls and ending child marriage in Tanzania.
        The World Bank, 2019 (Available at:)
        • Malhotra A.
        • Amin A.
        • Nanda P.
        Catalyzing gender norm change for adolescent sexual and reproductive health: Investing in interventions for structural change.
        J Adolesc Health. 2019; 64: S13-S15
        • Erulkar A.
        • Medhin G.
        • Weissman E.
        The impact and cost of child marriage prevention in rural Ethiopia.
        2017
        • Kennedy C.E.
        • Fonner V.A.
        • Armstrong K.A.
        • et al.
        The evidence project risk of bias tool: Assessing study rigor for both randomized and non-randomized intervention studies.
        Syst Rev. 2019; 8: 3
        • Centre for Population and Development Activities
        Adolescent girls in India choose a better future: An impact assessment.
        CEPDA, Washington, DC2001
        • Kanesathasan A.
        • Cardinal L.J.
        • Pearson E.
        • et al.
        Catalyzing change: Improving youth sexual and reproductive health through DISHA, an integrated program in India.
        Int Cent Res Women, Washington DC2008
        • Shahnaz R.
        • Karim R.
        Providing microfinance and social space to empower adolescent girls: An evaluation of BRASC’s ELA Centres.
        2008
        • Diop N.J.
        • Moreau A.
        • Benga H.
        Evaluation of the long-term impact of the TOSTAN programme on the abandonment of FGM/C and early marriage: Results from a qualitative study in Senegal.
        2008
        • UNICEF
        Long-term evaluation of the Tostan programme in Senegal: Kolda, Thiès and Fatick regions.
        Stat Monit Sect N Y UNICEF. 2008; 1: 1
        • Daniel E.E.
        • Nanda R.
        The effect of reproductive health communication interventions on age at marriage and first birth in rural Bihar, India: a retrospective study. Watertown: Pathfinder International.
        (Available at:)
        • Stark L.
        • Asghar K.
        • Seff I.
        • et al.
        Preventing violence against refugee adolescent girls: Findings from a cluster randomised controlled trial in Ethiopia.
        BMJ Glob Health. 2018; 3: e000825
        • Stark L.
        • Seff I.
        • Asghar K.
        • et al.
        Building caregivers’ emotional, parental and social support skills to prevent violence against adolescent girls: Findings from a cluster randomised controlled trial in Democratic Republic of Congo.
        BMJ Glob Health. 2018; 3: e000824
        • Bandiera O.
        • Buehren N.
        • Burgess R.
        • et al.
        Women’s empowerment in action: Evidence from a randomized control trial in Africa.
        World Bank, Washington, DC2018
        • Angrist J.
        • Bettinger E.
        • Bloom E.
        • et al.
        Vouchers for private schooling in Colombia: Evidence from a randomized natural experiment.
        Am Econ Rev. 2003; 92: 1535-1558
        • Pande R.
        • Kurz K.
        • Walia S.
        • et al.
        Improving the reproductive health of married and unmarried youth in India.
        2006
        • Gulemetova M.
        Evaluating the impact of conditional cash transfer programs on adolescent decisions about marriage and fertility: The case of Oportunidades.
        2011
        • Alam A.
        • Baez J.E.
        • Del Carpio X.V.
        Does cash for school influence young women’s behavior in the longer term? Evidence from Pakistan.
        The World Bank, Washington, DC2011
        • Jensen R.
        Do labor market opportunities affect young women’s work and family decisions? Experimental evidence from India.
        Q J Econ. 2012; 127: 753-792
        • Heath R.
        • Mobarak A.M.
        Manufacturing growth and the lives of Bangladeshi women.
        J Dev Econ. 2014; 115: 1-15
        • Hallfors D.D.
        • Cho H.
        • Rusakaniko S.
        • et al.
        The impact of school subsidies on HIV-related outcomes among adolescent female orphans.
        J Adolesc Health. 2015; 56: 79-84
        • Handa S.
        • Peterman A.
        • Huang C.
        • et al.
        Impact of the Kenya cash transfer for orphans and vulnerable children on early pregnancy and marriage of adolescent girls.
        Soc Sci Med. 2015; 141 ((Thirumurthy H.) Department of Health Policy and Management, UNC Gillings School of Global Public Health, United States): 36-45
        • Nanda P.
        • Datta N.
        • Das P.
        • et al.
        Making change with cash? Impact of a conditional cash transfer program on age of marriage in India.
        2016
        • Dake F.
        • Natali L.
        • Angeles G.
        • et al.
        Cash transfers, early marriage, and fertility in Malawi and Zambia.
        Stud Fam Plann. 2018; 49: 295-317
        • Koski A.
        • Strumpf E.C.
        • Kaufman J.S.
        • et al.
        The impact of eliminating primary school tuition fees on child marriage in Sub-Saharan Africa: A quasi-experimental evaluation of policy changes in 8 countries.
        PLoS One [Internet]. 2018; 13: e0197928
        • Hahn Y.
        • Islam A.
        • Nuzhat K.
        • et al.
        Education, marriage, and fertility: Long-term evidence from a female stipend program in Bangladesh.
        Econ Dev Cult Change. 2018; 66: 383-415
        • Duflo E.
        • Dupas P.
        • Kremer M.
        Education, HIV, and early fertility: Experimental evidence from Kenya.
        Am Econ Rev. 2015; 105: 2757-2797
        • Baird S.
        • McIntosh C.
        • Özler B.
        When the money runs out: Do cash transfers have sustained effects on human capital accumulation?.
        The World Bank, Washington, DC2016 (Policy research working Paper No. 7901)
        • Baird S.
        • Chirwa E.
        • McIntosh C.
        • Özler B.
        The short-term impacts of a schooling conditional cash transfer program on the sexual behavior of young women.
        Health Econ. 2010; 19: 55-68
        • Baird S.
        • McIntosh C.
        • Özler B.
        Cash or condition? Evidence from a cash transfer experiment.
        Q J Econ. 2011; 126: 1709-1753
        • Amin S.
        • Ahmed J.
        • Saha J.
        • et al.
        Delaying child marriage through community-based skills-development programs for girls: Results from a randomized controlled study in rural Bangladesh.
        2016
        • Amin S.
        • Saha J.S.
        • Ahmed J.A.
        Skills-building programs to reduce child marriage in Bangladesh: A randomized controlled trial.
        J Adolesc Health. 2018; 63: 293-300
        • Erulkar A.
        • Medhin G.
        • Weissman E.
        The impact and cost of child marriage prevention in rural Tanzania.
        2017
        • Erulkar A.
        • Medhin G.
        • Weissman E.
        The impact and cost of child marriage prevention in rural Burkina Faso.
        2017
        • Buchmann N.
        • Field E.
        • Glennerster R.
        • et al.
        Power vs money: Alternative approaches to reducing child marriage in Bangladesh, a randomized control trial.
        (Unpubl Manuscr)2018
        • Chae S.
        Timing of orphanhood, early sexual debut, and early marriage in four sub-Saharan African countries.
        Stud Fam Plann. 2013; 44: 123-146
        • Raj A.
        • McDougal L.
        • Silverman J.G.
        • Rusch M.L.A.
        Cross-Sectional time series analysis of associations between education and girl child marriage in Bangladesh, India, Nepal and Pakistan, 1991-2011.
        PLoS One [Internet]. 2014; 9: e106210
      2. Mathur S, Greene M, Malhotra A. Too Young to Wed: Education and Action Toward Ending Child Marriage. Washington, DC: International Center for Research on Women. 2003. Available at: https://www.icrw.org/wp-content/uploads/2016/10/Too-Young-to-Wed-Education-and-Action-Toward-Ending-Child-Marriage.pdf. Accessed September 20, 2020.

        • Paul P.
        Effects of education and poverty on the prevalence of girl child marriage in India: A district–level analysis.
        Child Youth Serv Rev. 2019; 100: 16-21
        • Glynn J.R.
        • Sunny B.S.
        • DeStavola B.
        • et al.
        Early school failure predicts teenage pregnancy and marriage: A large population-based cohort study in northern Malawi.
        PLoS One [Internet]. 2018; 13: e0196041

      Linked Article