There is substantial causal evidence documenting that later marriage yields better education and labor market outcomes for women (Field and Ambrus, 2008, Asadullah and Wahhaj, 2019, Sunder, 2019), as well as better health and education outcomes for their children (Sekhri and Debnath, 2014, Chari et al., 2017, Asadullah and Wahhaj, 2019, Garcia-Hombrados, 2022). Yet, many countries continue to document high rates of child marriage (UNICEF, 2017), for which the existing literature has explored a variety of economic and sociocultural explanations (Corno et al., 2020, Psaki et al., 2021, Buchmann et al., 2023, Corno and Voena, 2023). For governments interested in eliminating the practice of child marriage, age-of-marriage laws are, perhaps, the most direct policy lever available. While many studies have examined the effects of these laws on the outcomes of women (Bergstrom and Özler, 2023), there has been much less research on the effects of these laws on the health of the next generation.

In this paper, we estimate the intergenerational effect of child marriage bans, which set a legal minimum age of marriage to 18, on mortality rates among children of the affected women. Existing work shows that child marriage bans improve women’s socioeconomic outcomes and delay age of marriage in some contexts (McGavock, 2021, Rokicki, 2021, Wilson, 2022). We also know, primarily from studies that use age of menarche as an instrumental variable, that delaying age of marriage leads to better outcomes for children (Sekhri and Debnath, 2014, Chari et al., 2017, Asadullah and Wahhaj, 2019, Sunder, 2019, Garcia-Hombrados, 2022). However, it is not clear whether and to what extent child marriage bans will improve child health, especially given that age of marriage laws are often not properly enforced (Batyra and Pesando, 2021, Collin and Talbot, 2023) and can sometimes lead to substitution away from marriage to informal unions (Bellés-Obrero and Lombardi, 2023). In addition, child marriage bans may have the unintended effect of making child brides more hesitant to seek prenatal or postnatal care (for fear of legal punishment), which could lead to worse health outcomes for their children. In short, it is not clear whether child marriage bans will generate health benefits for the next generation.1

For this analysis, we use the MACHEquity Child Marriage Policy Database, which contains information on child marriage bans over the period 1995 to 2012, to identify the timing of the bans in each country (World Policy Center, 2022). This country-level dataset is then linked to the Demographic and Health Surveys (DHS), representative surveys of women aged 15–49 in low- and middle-income countries (LMICs), from which the main outcomes are extracted. Our final dataset includes 17 countries that legally banned child marriage under the age of 18 during the period from 1995 to 2012.

To isolate the causal effects of the bans, we rely on a similar strategy as Wilson (2022), which uses the same data to analyze the effect of child marriage bans on women’s socioeconomic outcomes. This strategy exploits two sources of variation: subnational regional differences in the pre-ban age at marriage and variation across cohorts within countries in exposure to the bans. We first calculate a region-specific measure of treatment intensity, defined such that locations where, in the pre-ban period, the occurrence of child marriage was common and child brides married particularly young are considered to have high treatment intensity, which means that individuals in these locations should be more affected by a child marriage ban.2 The treatment intensity variable is then interacted with an indicator for cohorts exposed to the bans (i.e., individuals under the age of 18 at the time a ban was implemented in their country). If the bans reduce child mortality, then there should be a larger decrease from women who turned 18 before to those who turned 18 after a ban in areas with high treatment intensity compared to areas of low treatment intensity.

Our results show that the bans have statistically significant and sizable effects on child mortality. Specifically, our main specification reveals a one standard deviation (SD) increase in treatment intensity reduced infant and under-5 mortality by about 0.95 and 1.97 percentage points, respectively (approximately 14.2 and 19.7 percent relative to the mean of pre-ban cohorts). Effects were driven by low-income countries. For under-5 mortality, the pattern of our coefficients from a standard event study analysis, as well as the approach proposed by Callaway et al. (2024), confirms our findings and lends support to the validity of our empirical strategy. For infant mortality, evidence from these alternative specifications is much weaker. Our main regression estimates are robust to various specification tests, including those that account for contemporaneous policies, differential trends due to regional characteristics, alternative samples, and migration. We thoroughly explore issues relating to selection into motherhood.

The bans appear to be reducing child mortality primarily by delaying age at first marriage and first birth. Women going into marriage or pregnancy with more maturity should have more agency and bargaining power, which are likely to be important for prenatal and postnatal health investments. Although this information is only available for a subset of our original sample, we document suggestive evidence that child marriage bans affect mothers’ health-seeking decisions around the time of birth (specifically, place of delivery and child vaccinations).

Our paper contributes to a large body of evidence on the impacts of child marriage laws on the outcomes of women and their children. While most work has focused on the impacts on women – age at marriage, education, and economic outcomes (Dahl, 2010, Bharadwaj, 2015, Wang and Wang, 2017, McGavock, 2021, Wilson, 2022, Bellés-Obrero and Lombardi, 2023, Collin and Talbot, 2023) – our study aims to shed light on the impacts on child mortality, for which existing evidence is scant, mixed, and currently only focused on Ethiopia (McGavock, 2021, Rokicki, 2021, Garcia-Hombrados, 2022). A key distinguishing feature of our study, compared to these three, is its scope: our sample includes 17 developing countries across four continents. In addition to providing more external validity, the use of pooled data from multiple countries helps reduce the possibility of biased estimates due to the low rates of infant and under-5 mortality, as discussed by Dursun et al. (2017).3 This also allows us to investigate sources of heterogeneity and potential underlying mechanisms.

This study is also closely related to a well-established literature on the effects of delaying marriage (outside of the context of child marriage laws). Studies in this literature typically use age at menarche as an instrumental variable for age at marriage (Field and Ambrus, 2008, Sekhri and Debnath, 2014, Chari et al., 2017, Asadullah and Wahhaj, 2019, Sunder, 2019, Garcia-Hombrados, 2022, Tauseef and Sufian, 2024). Because age at menarche may not satisfy the exclusion restriction for a number of reasons, our paper offers valuable evidence on this topic by using a different source of variation.4

Child marriage bans are just one example of an intervention aimed at reducing child marriage, and our study therefore speaks to a broader literature on this more general class of interventions. Bergstrom and Özler (2023) review studies on 15 categories of interventions (including age of marriage laws, job opportunities, and cash transfers) and conclude that none have been consistently found to be effective at reducing child marriage. Nevertheless, several recent papers have documented that community-based education and economic incentives can reduce child marriage (Chow and Vivalt, 2022, Buchmann et al., 2023, Cohen et al., 2023, Giacobino et al., 2024). The results of our study suggest that the benefits of these interventions could extend to the children of the affected women as well.