What Are the Drivers of Child Marriage? A Conceptual Framework to Guide Policies and Programs

Purpose: The child marriage field lacks a simplified framework that connects an understanding of the drivers of child marriage for girls to decisions about the design of interventions to delay marriage within different contexts and support married girls. Methods: We reviewed existing child marriage frameworks and conducted consultations with experts working on child marriage. We then developed a simplified conceptual framework describing the key drivers of child marriage for girls. We explored how these drivers play out and interact using qualitative data from three settings where child marriage is common: Bangladesh, Malawi, and Niger. Results: The final conceptual framework lays out five core drivers of child marriage for girls, which vary and interact across contexts. Social norms and poverty are shown as core drivers that underlie lack of agency, lack of opportunity, and pregnancy/fear of pregnancy. These drivers reflect community, household, and individuallevel factors. The case studies highlight the important relationships between these drivers, and the way they interact within each context. We use these examples to explore how policymakers and practitioners might identify the most appropriate interventions to address child marriage across different settings. Conclusions: We offer this framework as a starting point to guide more targeted interventions and policies that address the complex combination of child marriage drivers within each setting. By adapting this framework to different settings, thosedesigningand implementing childmarriageprevention interventions can identify thekey drivers in each setting, understand how those drivers interact, andmore effectively target effective interventions. 2021 Published by Elsevier Inc. on behalf of Society for Adolescent Health andMedicine. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Conflicts of interest: The authors have no conflicts of interest to disclose. Disclaimer: Publication of this article was funded by the Bill & Melinda Gates Foundation. The opinions or views expressed in this article are those of the authors and do not necessarily represent the official position of the funder. * Address correspondence to: Stephanie Psaki, Ph.D, 3939 N. Wakefield St., Arlington, VA 22207. E-mail address: spsaki@gmail.com (S.R. Psaki). 1054-139X/ 2021 Published by Elsevier Inc. on behalf of Society for Adolescent Health and Medicine. This is an open access article und creativecommons.org/licenses/by/4.0/). https://doi.org/10.1016/j.jadohealth.2021.09.001 According to UNICEF, among 20to 24-year-old women in least developed countries, 12% were married before the age of 15, and 38% before the age of 18 (2019; based on data from 2007 to 2017) [1]. Prevalence of child marriage is highest in parts of West and Central Africa, reaching 76% of girls in Niger [1]. In terms of burden, the largest number of girls affected by child marriage is in South Asia. UNICEF estimates that 25 million child marriages have been prevented over the last decade, but a substantial acceleration of progress is needed to eliminate the practice by 2030 [2]. Child marriage has similarities across settingsdmarriage before age 18 is illegal in many contexts and is often associated with low levels of education for girls, along with numerous er the CC BY license (http:// S.R. Psaki et al. / Journal of Adolescent Health 69 (2021) S13eS22 S14 negative health outcomes. Yet these experiences also diverge in important ways: in some settings child marriage may be driven largely by social norms, while in others poverty or unplanned pregnancy may be more important determinants [3e6]. As noted in a recent review of child marriage prevention interventions, cash transfers to support girls’ education appear to be promising [7]; however, this finding is based on a limited number of interventions and insufficient geographic representation, so it remains difficult to pin down which interventions will be most effective in each context [7]. Researchers have proposed various frameworks laying out the hypothesized drivers, correlates, and outcomes of child marriage, with a good deal of overlap between frameworks. Yet, many of these frameworks are complex, narrowly focused on one setting, or they lack a clear connection to policies or interventions. As a result, the field lacks a simple, unified framework that brings together an understanding of the drivers of child marriage for girls with clear implications for how programs and policies can better target the unique combination of drivers between and within settings. We draw on previous research to develop a streamlined and practical framework of the drivers of child marriage across and within settings. This simplified framework, which draws on a review of the literature, expert consultations, and case studies in three countries, proposes five interrelated drivers of child marriage: poverty and economic factors; lack of opportunity; social norms and attitudes; lack of agency; and fear of girls’ sexuality and pregnancy. As the child marriage prevention field continues to build the evidence base on how best to delay marriage across settings, this framework can serve as a valuable tool in identifying the most appropriate policies or interventions to eliminate child marriage in each setting. Overview of approach The initial work to develop this streamlined conceptual framework was completed in 2020 after consultation with a consortium of organizations working to address child marriage, including researchers, advocates, practitioners, and donors. We began by conducting a review of the empirical literature on potential drivers of child marriage, including a review of previous frameworks, some of which were developed by members of the consortium. We also engaged in expert consultations with individuals working on child marriage prevention from research, policy, and advocacy perspectives. Based on that process, we developed a draft simplified conceptual framework. Using qualitative data from Bangladesh, Malawi, and Niger, we further explored the intersections between drivers of child marriage through case studies. Based on the results of case studies, we made several small modifications to the framework. We begin by summarizing the literature review and sharing the final conceptual framework. We then present the three case studies, which demonstrate how the drivers of child marriage play out across and within each context. We then reflect on how this framework might be useful to inform future policies and interventions designed to prevent child marriage. This work was approved as exempt by the Population Council Institutional Review Board as it includes secondary analyses of deidentified data. The original data collection was also conducted by the Population Council and all approved by the Population Council Institutional Review Board. Informed consent procedures included parental permission and adolescent assent (for minors), and informed consent for adults 18þ and emancipated minors.

negative health outcomes. Yet these experiences also diverge in important ways: in some settings child marriage may be driven largely by social norms, while in others poverty or unplanned pregnancy may be more important determinants [3e6]. As noted in a recent review of child marriage prevention interventions, cash transfers to support girls' education appear to be promising [7]; however, this finding is based on a limited number of interventions and insufficient geographic representation, so it remains difficult to pin down which interventions will be most effective in each context [7]. Researchers have proposed various frameworks laying out the hypothesized drivers, correlates, and outcomes of child marriage, with a good deal of overlap between frameworks. Yet, many of these frameworks are complex, narrowly focused on one setting, or they lack a clear connection to policies or interventions. As a result, the field lacks a simple, unified framework that brings together an understanding of the drivers of child marriage for girls with clear implications for how programs and policies can better target the unique combination of drivers between and within settings.
We draw on previous research to develop a streamlined and practical framework of the drivers of child marriage across and within settings. This simplified framework, which draws on a review of the literature, expert consultations, and case studies in three countries, proposes five interrelated drivers of child marriage: poverty and economic factors; lack of opportunity; social norms and attitudes; lack of agency; and fear of girls' sexuality and pregnancy. As the child marriage prevention field continues to build the evidence base on how best to delay marriage across settings, this framework can serve as a valuable tool in identifying the most appropriate policies or interventions to eliminate child marriage in each setting.

Overview of approach
The initial work to develop this streamlined conceptual framework was completed in 2020 after consultation with a consortium of organizations working to address child marriage, including researchers, advocates, practitioners, and donors. We began by conducting a review of the empirical literature on potential drivers of child marriage, including a review of previous frameworks, some of which were developed by members of the consortium. We also engaged in expert consultations with individuals working on child marriage prevention from research, policy, and advocacy perspectives. Based on that process, we developed a draft simplified conceptual framework. Using qualitative data from Bangladesh, Malawi, and Niger, we further explored the intersections between drivers of child marriage through case studies. Based on the results of case studies, we made several small modifications to the framework.
We begin by summarizing the literature review and sharing the final conceptual framework. We then present the three case studies, which demonstrate how the drivers of child marriage play out across and within each context. We then reflect on how this framework might be useful to inform future policies and interventions designed to prevent child marriage.
This work was approved as exempt by the Population Council Institutional Review Board as it includes secondary analyses of deidentified data. The original data collection was also conducted by the Population Council and all approved by the Population Council Institutional Review Board. Informed consent procedures included parental permission and adolescent assent (for minors), and informed consent for adults 18þ and emancipated minors.

Poverty and economic factors
Numerous studies have found an association between household socioeconomic status and age at marriage (e.g., in Bangladesh [8e10]; in India [11]; in Ghana [6]), with poverty closely linked to lower age at marriage. Although this relationship is fairly consistent, with the poorest girls being at highest risk of child marriage in almost every country where the practice is prevalent, the mechanisms through which this relationship operates vary between settings. Child marriage might be an economic strategy to reduce the financial burden on families of caring for or educating daughters, especially in communities with limited opportunities for female labor force participation. Girls may choose to marry when their parents are unable to provide for them at home, or to support them in continuing in school. Finally, marriage transactions, including transfers from the bride's family to the groom's or vice versa [12], are often a key element of decisions about the timing of marriage. However, the relationship between poverty and marriage timing may depend in part on the nature of the marriage market in each setting [13e15].

Lack of opportunity
Closely linked with poverty, girls and their families may see marriage as the only viable pathway in settings where opportunities for continued schooling, labor force participation, or other valued roles are limited. Most immediately, growing evidence shows that access to schooling has the potential to delay marriage for girls, in part because school enrollment is considered incompatible with marriage and childbearing in many settings. Evidence shows that those who marry before the age of 18 tend to complete fewer years of schooling (e.g., in West Africa [16]; in Nepal [17]; in India [18]; in Malawi [19]; in Kenya, Senegal, Uganda, Zambia [20]; in Ethiopia [21]). Yet in some settings this link between marriage and schooling may not be as strong as assumed, in part due to poor school quality, which results in weak skills, leading girls and their parents to question the value of girls' education [22]. That is, even when education opportunities are available, in settings where women are unable to translate education into longer term opportunities, and/or where social norms discourage women's employment, the effects of education opportunities on delaying marriage may be limited. Lack of labor market opportunities for women, especially those requiring academic skills, may reinforce marriage as the only viable pathway for girls and their families, even when the cost of education is minimal [12,20,23e25]. Aside from education and employment opportunities, girls in some settings may view marriage as the only pathway to independence from their parents (in Tanzania [5]; in Iran [26]).

Social norms and attitudes
A fair amount of attention has been paid in the theoretical and empirical literature to the effects of social norms, especially gender norms, as drivers of child marriage. Bicchieri defines a social norm as, "a collective practice sustained by empirical and normative expectations and by preferences conditional on both these expectations" [27]. As Greene and Stiefvater [28] point out, social norms connected with child marriage cover a range of domains, including the transition to adulthood, sexuality, age hierarchies, religious beliefs, gender inequality, and women's and men's economic roles. This literature distinguishes between social expectations, related to what others believe and do, and nonsocial expectations, or personal beliefs that are not dependent on what others believe. If parents' preferences for child marriage are informed, in part, by the belief that child marriage will help protect their daughters, then sharing information about the potential harms of child marriage may be sufficient to drive change. On the other hand, if child marriage decisions are largely driven by social expectations, then more concerted community action may be needed. The norms that drive child marriage may reflect broader gender norms, including those related to sexuality, mobility, and/or participation in the labor market. In practice, a combination of personal beliefs and social expectations may drive child marriage decisions in many settings.
Numerous studies have shown that child marriage is linked with parents' attitudes and community norms in support of child marriage (e.g., in Indonesia [24]; in Niger [29]; in Ghana [6]; in India [30]). Aligned with theory on social norms, evidence shows that community norms in support of child marriage often reflect parents' (and their daughters) normative beliefs, which perhaps reflect internalized social norms, rather than merely social pressure to marry (in India [30]; in Pakistan [31]; in Ethiopia [32]; in Ghana [6]). In some contexts, girls may express a desire to marry [26,33,34], perhaps reflecting social norms in favor of marriage and the lack of viable alternatives. The prevalence of child marriage is often erroneously equated with social expectations of child marriage specifically, when other norms around girls' education, women's employment, or girls' sexuality may be more relevant (in Senegal, Zambia, Uganda, Kenya [20]; in Malawi [35]). For example, Steinhaus et al. [35] found that, despite lack of social norms in favor of child marriage, this practice was common due to unplanned pregnancies among unmarried girls, and norms in favor of marriage once a pregnancy occurs. In short, social norms may create pressure to marry despite household preferences, or they may be aligned with parents' or girls' attitudes in support of child marriage. Norms may also be closely connected to other drivers of child marriage, such as lack of opportunity or agency and fear of pregnancy.

Lack of agency
In situations where parents or girls wish to delay marriage, whether or not the marriage occurs may depend on agency, or "the ability to define one's goals and act on them" [36]. In some settings girls may have little control over the decision to marry, perhaps reflecting strong social norms in support of child marriage (e.g., in Bangladesh [37]; in Indonesia [38]). For girls, both identifying the goal of delaying marriage and acting on it may present challenges, especially in settings with limited information about the consequences of child marriage and limited availability of alternative pathways. Or, girls may lack agency in the decision to marry in certain circumstances, such as if a pregnancy occurs before marriage. Within the same setting, some girls may have more agency than others (in Ethiopia [39]; in India and Ethiopia [40]), and some may not even know that a marriage is planned in advance [41], making agency less relevant to the marriage process. As noted previously, girls may also use their agency to choose to get married in settings where few alternatives exist, or where social pressure to conform with child marriage norms is strong (in Honduras [42]; in Somaliland and Puntland [43]; in Tanzania [44]; in Brazil [45]). When girls do express their agency to resist marriage, they may be forced to leave their homes anyway, as is the case for those who migrate to avoid child marriage [46].

Fear of girls' sexuality and pregnancy
In communities where patriarchal gender norms are dominant, and girls' virginity is highly valued, families and communities may support child marriage as both the only legitimate context of sexual activity for girls, as well as protection against potential disgrace if girls circumvent those norms [47]. Religious beliefs and practices may also encourage child marriage to protect girls from premarital sex and ensure religious norms and expectations around virginity and purity are upheld [48]. Child marriage may be a strategy used by parents to protect their families and daughters from perceived threats, such as concerns about family honor, girls' virginity, sexual assault, or pregnancy that occurs outside of marriage (in Tanzania [5]; in Ghana [6]). Or, child marriage may occur in response to pregnancy among unmarried girls, in settings where premarital pregnancy is seen as a greater taboo than child marriage (in Senegal, Zambia, Uganda, Kenya [20]; in Zambia [49]; in Malawi [35]). For example, Stark [5] found that parents of adolescent unmarried girls in Tanzania see marriage as a way to protect their daughters from transactional sex, a practice largely driven by poverty, as well as a potential pregnancy that might result. In Malawi, 85% of adult "decision-makers" reported that child marriage happened as a result of pregnancy, even though the practice was not endorsed by most as desirable [35]. Studies in other settings have also found that earlier age at menarche is associated with younger age at marriage (in Bangladesh [9]; in Indonesia [24]), potentially as a strategy to avoid pregnancy outside of marriage in more conservative settings.

How interventions address the drivers of child marriage
Several reviews of interventions designed to address child marriage have been conducted over the last decade, which divide approaches into several broad categories [7,50,51]: 1. empowerment programs, which aim to increase girls' agency and equip them with knowledge and skills to avoid child marriage; 2. community engagement programs, which aim to address social norms by sensitizing parents and community members to the risks of child marriage; 3. education interventions, which encourage support for continued education as an alternative to marriage; 4. economic support programs, which aim to alleviate economic pressures and offer financial incentives for certain behaviors (e.g., delaying marriage, keeping girls in school); and 5. legal or policy interventions, which aim to create a legal or policy environment that makes child marriage more difficult.
However, recent reviews have drawn somewhat different conclusions about which components are most important for delaying marriage. Authors of a 2012 review argued that the most effective child marriage prevention programs were those that combined empowerment components with community engagement, but raised questions about scalability, selectivity of participants, and the sustainability of changes [50]. In a 2016 review, Kalamar et al. [51] identified four examples of effective interventions, three of which provided an economic incentive for households to keep their daughters unmarried, while the fourth focused on building life skills. However, the authors also identified evaluations of similar economic approaches that found no effects. Based on results from 14 rigorously evaluated interventions, Chae and Ngo argued that empowerment content is an important component of programs designed to delay marriage in effective studies. However, only 57% of programs that included an empowerment program were successful, indicating that more information is needed on the circumstances in which this approach is most effective [52]. Most recently, Malhotra and Elnakib find that cash transfers for education are the most promising intervention and argue that they operate by increasing girls' human capital. Although Malhotra and Elnakib's [7] review is the most recent, there is still a need for clarity on which intervention approaches are most effective in each setting as the evidence base is not geographically comprehensive.
Taken together, existing evidence on the effects of interventions underlines the fact that drivers of child marriage vary between, and even within, settings, and interventions that work in one setting may be ineffective, or less cost-effective, in another setting. Therefore, a clearer understanding of the drivers of child marriage within and between settings can inform the design of more effective, and more cost-effective, interventions to delay marriage and expand opportunities for adolescent girls.

Conceptual framework
In addition to findings from available theoretical and empirical literature, we draw on previous conceptual frameworks describing the drivers or consequences of child marriage in different contexts. Although we are unable to capture all previous work due to space constraints, we highlight selected examples here, connecting each to our new proposed framework. Several previous frameworks explore the ways in which decisions about child marriage are influenced by factors at the individual, household, and community levels. For example, UNICEF and UNFPA [53] describe the influence of macro socioeconomic factors (e.g., levels of women's empowerment, demographic changes, economic situation) on individual and household-level decisions about marriage in South Asia (Bangladesh, Nepal, Pakistan). Many frameworks also explore the relative importance of economic versus sociocultural drivers of child marriage, even while pointing out that the distinctions may not be as clear in practice as they are in theory [54]. In contrast, Jones et al. [55] found that poverty did not appear to be an important driver of child marriage in some parts of Ethiopia, although a lack of opportunity (linked to poverty) did appear to play a role. Several previous studies point to the complexity of the relationship between agency/decision-making and child marriage, describing the relationship as a "double-edged sword" [55]. For example, Taylor et al. [56] used qualitative data from three countries in Latin America (Brazil, Guatemala, Honduras) to explore the ways different types of girls' and young women's agency may play out either to drive child marriage (oppositional agency, accommodating agency) or to delay marriage (transformative agency). Although a few studies (e.g., in Ethiopia [54,55]) explore the connections between drivers of marriage and common intervention approaches or changing patterns of opportunities for girls, most existing frameworks describe the factors leading to child marriage without connecting them to the design and implementation of effective interventions [53].
A theory of change laid out by Girls Not Brides, with input from a range of organizations, presents a broad overview of the vision of a world without child marriage, and describes four programmatic strategies to address child marriage: empowering girls, mobilizing families and communities, providing services, and establishing and implementing laws [57]. This broad vision is an essential starting point for discussions on how to delay marriage. We aim to build on that work by simplifying a set of common drivers and linking those drivers to common intervention approaches.
Enormous progress has been made in recent years in addressing child marriagedboth in terms of building an evidence base, and in achieving declines in the proportion of women married before the age of 18. And yet, policymakers, program implementers, and advocates have called for more "clarity on strategic and cost-effective interventions" to accelerate progress and achieve the Sustainable Development Goals [58]. Within the child marriage field, as with many other areas, building the evidence base may be insufficient to drive clear, consistent progress unless that evidence is synthesized and translated into simple, actionable tools that can be used by practitioners. Figure 1 shows a conceptual framework laying out the hypothesized drivers of child marriage that emerged most consistently from a review of the literature, including previous conceptual frameworks or theories of change, and expert consultations. Social norms and attitudes and poverty and economic factors are shown as core distal conditions that in turn drive more proximal factors leading to child marriage. Lack of agency, lack of opportunity, and fear of girls' sexuality and pregnancy reflect community, household, and individual-level factors. The large circle in the middle represents the important connections between these five drivers, which are likely to intersect for many girls and their families. Interventions and policies may target the drivers themselves, or the relationship between each driver and age of marriage (e.g., how fear of pregnancy affects risk of child marriage). The drivers of child marriage may also influence short, medium, and longer-term outcomes for girls and their families.
Although this conceptual framework draws on existing theory and evidence, interventions and policies are often designed and implemented without explicit attention to which drivers prevail in each setting due to lack of time, expertise, or data, among other reasons. We offer this framework, which can be adapted based on the drivers in each setting, as a tool to guide decision-makers in making the most effective investments to address child marriage. Furthermore, this framework can be leveraged by researchers to help guide discussions about the nature of child marriage across settings, and the evidence for which intervention approaches (e.g., empowerment programs vs. cash transfers vs. other) are most likely to be effective in different contexts.

Case Studies
This section presents three case studies drawing on qualitative data from Bangladesh, Niger, and Malawi to explore how the drivers of child marriage shown in Figure 1 interact in different settings. Data from these countries were chosen because of high levels of child marriage, varying levels of evidence on effective interventions, and the availability of rich qualitative data collection from a recent child marriage intervention program, including specific questions on perceived drivers. The conceptual framework presents a simplified set of five drivers that are closely connected. However, as is often the case, in practice these relationships are more complex. The case studies highlight the important relationships between these drivers, and the ways they interact in each context. Table 1 summarizes the data and methods used in each case study.
Case study 1: girls' agency, social norms, lack of opportunity, and poverty in the context of marital choice in Bangladesh In Case Study 1, we explore questions about the role of agency in marriage decisions, and its connection to other drivers, using qualitative data from the BALIKA study, conducted in rural Bangladesh (Table 1). Although poverty and limited economic opportunities are important underlying drivers of child marriage in this setting, village norms and customs supporting arranged marriages, which grant girls little or no agency in their marriage decision-making, are primary drivers. Poverty provides important context for these decisions. Child marriage reduces the economic burden of providing for adolescent girls and may be a stronger motivation in poor families. Fear that premarital sex may affect a girl's prospects for a marriage partner may put added pressure on the decision. In this context, girls remain passive actors, sometimes going along with their families' decisions to marry them rather than resisting even if they are not ready for marriage.
Interviews conducted postintervention explored whether views and attitudes about marriage had shifted. Girls believed that they have little agency in decisions about marriage timing or choice of partner. Despite the fact that more than 90% of marriages are registered, a process that requires girls signing a "nikah namah" to signal their consent, most married girls said they were never consulted before their marriages. Although girls expressed disappointment at their lack of agency in marriage decisions, they also provided reasons they do not challenge the status quo: they do not have decision-making power, conforming to their families' decisions is in their best interest in the long run, and financial considerations. I: Why were you married off? P1: The proposal came from a relative . the groom's family was uneducated but somewhat solvent. My parents agreed to the proposal hastily and I could not disobey and agreed to marry.
Girls described several factors that drive the timing of marriage: reaching puberty and related fears about girls' safety and security, fear of scandals or rumors, girls' beauty, receiving a marriage proposal that is "too good to refuse," and peer and community pressure. Girls described parents' fears about their daughters' reputations after reaching puberty, especially if they learned that other men were expressing interest, due to concerns about the loss of family honor if girls engage in "love marriages." Girls explained that such marriages dishonor the family and are only chosen by disobedient girls who are unwilling to follow social norms. The large circle in the middle represents the important interactions between these five drivers though we note the arrows could go in multiple directionsdfor the sake of simplification we only note that the larger social context influences agency, opportunity, and fear of girls' sexuality/pregnancy. Interventions and policies may target the drivers themselves, or the relationship between each driver and age of marriage. The drivers of child marriage may also influence short, medium, and longer term outcomes for girls and their families. P2: If a girl is not married after her puberty, people start talking about her character. The parents are taunted by neighbors, so they prefer to give the daughters in marriages instead.
P3: Some girls are also a little outgoing type. Parents are not able to control them. So parents think it's better to marry them off, otherwise they would bring bad reputation to the family.
Finally, girls also expressed that limited alternatives exist to marriage due to both their families' financial constraints and the lack of opportunity for girls living in rural communities to contribute financially to their households. However, some girls expressed that continuing their education was considered a valid reason to delay marriage. P4: If a girl is poor and unable to continue schooling or to work .Then it is just better that she gets married.
These findings from girls' perspectives demonstrate that, although girls experience little agency in decisions about marriage, they also acknowledge the strong role of social norms in support of child marriage, and the lack of viable alternatives for girls living in rural communities.
In similar contexts, interventions that offer skills and knowledge to expand education and livelihood opportunities can potentially delay marriage if families see girls as an asset and not a burden. For such interventions to bring about change it is important for them to be visible to the community and perceived positively. Interventions that lead to changes in community norms that empower girls and privilege their voices so that they can exercise their agency in their marriage timing and decision may be effective.
Case study 2: agency, social norms, poverty, lack of opportunity, and child marriage in Niger In contrast to the Bangladeshi context, in Niger, high proportions of girls report that they have agency regarding marriage, with many married girls reporting that the decision to marry was their own choice [60e62]. Qualitative research conducted as a part of the More Than Brides Alliance (MTBA) study in Niger (Table 1) illuminates key features of the intersecting economic and normative contexts, which help to explain why adolescent girls seem to hold largely supportive attitudes toward child marriage in the country with the highest prevalence of child marriage in the world.
In Niger, economic and normative factors interact to influence girls' attitudes toward marriage and the timing of their marriages. The Nigerien context is characterized by extreme poverty and severely limited educational and economic opportunities, especially for girls and women, as well as by a normative environment in which labor is strictly divided by gender and there are clear expectations for the respective economic and social roles of women and men. Focus group participants indicated that working to bring money into the household is expected of men and is only considered appropriate for women in the case of necessity when one's husband cannot provide. Such attitudes about appropriate economic roles for girls and women likely contribute to the deprioritization of girls' education: the most recent DHS data show that approximately 70% of girls and 55% of boys ages 10e14 were not in school [62]. Consequently, girls and women remain largely financially dependent on men; there are few examples of women who have followed a livelihood strategy other than marriage that might serve as role models for girls in the community, and there exists little to no viable space for girls and women outside of their natal or marital homes. In this context, adolescent girls perceive economic dependence on husbands as preferable to dependence on parents, and marriage as a shift to adulthood and relative autonomy.
We would like to get married because we have reached the marriage age and sometimes we feel ourselves that we need a husband.

Unmarried girl, Tillabéri
Additionally, social norms and cultural practices contribute to the perpetuation of child marriage in Niger, namely the practice of marrying in cohorts of peers and the important emphasis placed on obedience as a value that is key to social harmony. We find that girls in the same peer group are married in cohorts, during a specific season of the year following the harvest. Within this system, there is a narrow window of time in girls' lives during which is it seen as acceptable to marry. Participants in this study stressed the benefits of marrying at the "right" time, and there was an overwhelming consensus among respondent groups that it is preferable for a girl to be married first rather than last among her peers.
I: Thinking about your community, are there examples of girls who do not marry at the same time as other girls who are their age? P: We cannot have that here in our village.
Mother, Maradi P: There are no girls of the same age of which some are married while others are there waiting.

Unmarried Girl, Maradi
The cultural context in Niger places high value on conformity and obedience for everyone, but especially for women and girls. A girl's decision to accept a marriage proposal not only reflects her conformity to others' expectations, but also signals more broadly that she will be an obedient and virtuous wife, daughterin-law, and adult member of the community. Furthermore, girls who are perceived as disobedient may be subject to social sanctions in their families and communities. P: If the girl marries the one that her parents do not like, she will see the consequences and will suffer them alone.

Mother, Tillabéri
Given the constrained economic opportunities and the norms in this context, there appears to be very little incentive for delaying marriage and few girls choose to go against the norm. Interventions in Niger should engage with families and communities to change perceptions of what roles are acceptable for girls beyond marriage and motherhood, encourage girls and their families to prioritize girls' education, and expand livelihood opportunities for girls and women. These findings from Niger demonstrate the dual challenge interventions face of: (1) improving material conditions to increase opportunities for girls and enable them to develop economic independence and (2) changing social norms related to the desirability of education and economic engagement by women and girls.
Case study 3: pregnancy, fear of girls' sexuality, social norms, and child marriage in southern Malawi Using qualitative data from the MTBA study, we explored the drivers of child marriage in Malawi with adolescents and their parents in Mangochi (Table 1). Girls report that pregnancy often precipitates marriage due to parents' expectations that girls will marry the biological father.
When they start having sex while they are young, they end up getting pregnant as a result their parents force them to get married.

FGD unmarried adolescents
The knowledge that pregnancy will often lead to marriage may influence partner choice for adolescent girls. As a result, the selection of a romantic partner may be the time when girls have the most agency in marital choice.
These girls are selecting the man that they think when they have become pregnant by him, he will not be able to deny her. Or even his parents will welcome her even if he impregnates her.

FGD unmarried adolescents
We also explored social norms around pregnancy and child marriage. Trained community members were asked to observe and record girls' behaviors and related community discussions based on a technique developed by Watkins and Swidler [63] in their research on HIV. Community members were trained to document public observations over several months, with a focus on child marriage and adolescent sexuality, and training included discussion of ethics consideration and guidance on protecting individuals' identities.
We found that community observers described adolescent pregnancies and marriages as consequences of girls' bad decisions. In their notes, community observers themselves also attributed the problem of child marriage to girls' behavior. This observation recalls a conversation: Mercy's (a pseudonym) immoral behavior is everywhere in the area and surrounded areas, he said. When she goes to play during the evening she comes back very late and other times she does not return home. Mercy does not listen to the advice given to her by either friends or her parents.

Community observer
He added that girls are not trusted nowadays. In addition to that his wife explained that this is the behavior of many girls. Many of them do not listen to the advice especially when they reach puberty.

Community observer
Girls report that sexual activity and marriage are inevitable and there is little incentive to delay marriage, especially if your choice of partner may become limited as peers begin to marry the most desirable partners. Yet communities view girls as irresponsible when they engage in sexual relationships and expect marriage if a pregnancy occurs. Although girls and community members describe pregnancy as a common driver of child marriage, more indirectly girls describe the roles of poverty and lack of opportunity in driving their choice of partner and decisions about sexual behavior. Interventions to address poverty (such as those by Baird et al. [64]) and livelihood interventions to provide alternatives to supporting oneself outside of marriage should be combined with social norms approaches to acknowledge girls' sexuality and offer accessible and supportive SRHR services to encourage delaying pregnancy.

Discussion
This paper presents a conceptual framework laying out five hypothesized drivers of child marriage: poverty/material context, social norms/normative context, lack of opportunity, lack of agency, and fear of girls' sexuality and pregnancy. The three case studies provide support for the framework and highlight the specific ways these drivers interact in different contexts to drive child marriage.
Overall, the three case studies described the different roles and connections between the hypothesized drivers of child marriage presented in our conceptual framework. Poverty/economic factors and social norms/attitudes contribute to a lack of opportunity for girls in many settings, which might make child marriage more appealing, both to girls and their families. As a result, girls may see child marriage as the best available option, such that, even when they do have agency in the marriage decision, their expression of agency may not always lead to later marriage. Social norms, particularly around fear of girls' sexuality, also emerge as closely tied with norms around marriage, pregnancy, and fear of pregnancy. In some settings this may lead to marriage in response to pregnancy, while in other settings parents might feel pressure to marry their daughters before a pregnancy occurs. The connection between social norms around girls' sexuality and pregnancy emerged from qualitative data in southern Malawi. Both adolescent girls and adults described a pattern of pregnancies among unmarried girls leading to child marriage, a reality that perhaps informs the choice of sexual partners by girls. Rather than strong social norms in support of child marriage, observers instead described disapproval of sex before marriage, consistent with other research in the region [35,49]. In contrast, qualitative data from Niger described strong norms in favor of child marriage, a view that appears closely linked with widespread poverty and lack of opportunities available to girls outside of marriage. Girls in Niger described their own support for child marriage, which they saw as a welcome move toward independence from their parents in a setting where few alternatives exist.
These case studies demonstrate shared drivers of child marriage across diverse settings, as well as important differences in how those drivers operate and interact. This diversity underscores the importance of moving beyond simple categorizations of settings based on levels of child marriage, to reflect the diversity of experiences more accurately, thereby developing more targeted and effective solutions. For example, while evaluations have shown that offering financial incentives to delay marriage is effective in certain settings (e.g., in Bangladesh [65]; in Malawi [64]; in Uganda [66]; in Ethiopia and Tanzania [67]; for a review [51]), Amin et al. [68] point out that traditional evaluations of these programs may fail to capture whether incentives to delay marriage translate into greater agency for girls and women within marriage. Previous research has also questioned the durability of these effects once the transfers end [64]. Other evaluations have demonstrated the limits of financial incentives in settings where few alternative pathways exist. Erulkar and Muthengi [69] report on the effects of an intervention in Ethiopia aiming to reduce child marriage through group formation, incentives for girls to remain in school, and community awareness raising. Although the program was effective at keeping younger girls (ages 10e14) unmarried and in school, older girls (ages 15e 19) who participated in the program had a higher risk of marriage by end line, indicating that alternative pathways beyond primary schooling were needed. Although the most recent review by Malhotra and Elnakib [7] notes that economic incentives to support girls' education may be the most effective intervention, it is unclear whether this approach would work in Niger, for example, where primary school enrollment is low and returns on education limited [70], though we note recent research has shown that CCTs can have an impact on school enrollment and delayed marriage in some lower enrollment contexts, such as northern Kenya [71].
The framework provides guidance on how programs to address child marriage should consider these drivers in designing interventions. In the framework we suggest that poverty/economic factors and social norms/attitudes are part of the broader environment in which decisions around child marriage are made, and these influence girls' ability to express agency, the opportunities that exist for girls and women beyond marriage, and how girls' sexuality is viewed. Similar to recommendations from the Child Marriage Research Network [59], this framework supports thinking about multilevel and intergenerational program approaches that address underlying power structures as part of the social context. For example, programs that aim to expand economic opportunities for girls should examine the social context in which economic opportunities exist. In Niger we suggest that this requires engaging with families and communities to change perceptions of what roles are acceptable for girls beyond marriage and motherhood and perhaps working on elevating mentors and role models. This approach is similar to the Gender Action Learning System developed by Oxfam [72], which demonstrated the impact on girls' work experience. In Malawi, this approach may mean addressing livelihood opportunities alongside acknowledging girls' sexuality and agency in sexual partner choice and the role of financial support in romantic relationships [73].
The framework may be helpful to stimulate thinking and discussions around the normative and economic context in each setting, and how those factors influence what agency and opportunity look like for girls in that context. We envision that decision-makers would use this framework in connection with available local data on child marriage and adolescent health. For example, recent survey data, such as those collected by the DHS program, would be useful to understand whether premarital sex is common, and the availability of education and livelihood opportunities for girls and women. Available local qualitative data may provide insight into social norms related to child marriage, as well as insight into girls' agency in decisions about marriage timing. Together, these data could help decision-makers understand which components may be most important in each setting. They could then look to recent evidence reviews to identify the most effective interventions to address relevant drivers, and to see whether recent evaluations have been conducted in similar contexts. For example, in the evaluations that provide the data for these case studies, interventions were adapted to each context. The MTBA was a five-country intervention, but was tailored in each country based on available data (e.g., schoolbased interventions would miss large groups of out-of-school girls in Niger; sexual and reproductive health and rightsfocused intervention approaches in India were not considered acceptable in Pakistan). Although complete data on drivers were not available when the interventions were designed, new information on drivers from the evaluation helped researchers to make sense of mixed findings for the MTBA evaluations.
There are a number of important limitations to our conceptual framework. First, the framework assumes availability of recent datadboth quantitative and qualitativedon key drivers in each context. Although ideal data are rarely availably, policymakers and practitioners may be able to fill gaps in evidence through more informal discussions with community leaders or girls themselves, including through intentional design approaches developed for these circumstances. Additionally, this framework does not explicitly incorporate legal or policy changes, due to both lack of evidence on their effects and an assumption that they often reflect changing norms. Third, while we assume many of the drivers of child marriage are exacerbated during times of crisis and displacement, the framework mostly draws on research conducted in relatively stable settings. Finally, this framework is focused on understanding the drivers of child marriage for girls due to the relatively higher prevalence, and might be less relevant in contexts where child marriage is common for boys.
We offer this framework as a starting point to guide more targeted interventions and policies that address the complex combination of child marriage drivers within each community. Rather than requiring complex data, this framework acknowledges that policymakers and practitioners are often operating with substantial constraints and might use a simplified framework to drive more systematic and targeted decisions. By adapting this framework to different settings, those designing and implementing child marriage prevention interventions can identify the key local drivers, understand how those drivers interact, and more effectively target policies and interventions.