Research

Child marriage remains widespread, although most countries have set a legal minimum marriage age of 18. Previous studies find that minimum-age reforms rarely delay marriage on average, but evidence on how their effects vary across contexts and how they affect women’s life-cycle outcomes remains limited. I study child marriage bans (CMBs), which set 18 as the minimum marriage age without parental-consent exceptions, using a regression discontinuity design based on birth-month eligibility. Combining WORLD policy data with Demographic and Health Surveys, I cover 20 countries that adopted a CMB between 1995 and 2023, the largest sample to date. While CMBs do not reduce child marriage on average, they can do so where prevalence was already low, with suggestive reductions where informal unions were less common, where the previous legal age was close to 18, and among girls aged 16 and 17. Independently of marriage timing, CMBs lead women to marry more educated husbands and reduce their justification of intimate partner violence. CMBs thus do not by themselves overturn established marriage practices but can reinforce social change already underway.

Significant resources are allocated to targeted aid projects aimed at enhancing the socio-economic well-being of participants. Evaluating these projects is both time-consuming and costly. A major difficulty in these evaluations is identifying an adequate control group to accurately measure the treatment effects of the aid projects. We propose a cost-effective method for selecting a control group that can serve as a baseline for post-project evaluations. Specifically, we explore the feasibility of using existing secondary data sources, such as household surveys, to establish a control group that mirrors the treatment group in evaluation analyses based on interviews. We validate our approach using data from an educational programme in rural Ghana that targeted marginalized, out-of-school adolescent women. This setting exemplifies a highly specific aid intervention, common in many small-scale development initiatives. Our findings highlight the critical role of considering the project’s selection criteria when constructing a control group from existing secondary data. Furthermore, we emphasize the necessity of ensuring that survey questions administered to the treatment group match those used in the secondary data source to maintain comparability in evaluation analyses.

Over the past decade, a substantial body of research has emerged leveraging large-scale educational policy reforms in Africa - so-called Free Primary Education (FPE) policies - to examine the causal effect of education on women’s sexual and reproductive health (SRH). Yet, reported effects are generally small and vary sufficiently such that zero mean effects cannot be ruled out. Importantly, part of this variability may reflect differences in sample construction and specification choices across studies, despite a strong concentration on only a small number of FPE countries. This paper offers a systematic, harmonised, and continent-wide quantitative evaluation of FPE-induced education on women’s SRH. Pooling all available Demographic and Health Survey (DHS) data from all countries that have ever implemented FPE, we employ best-practice, data-driven regression discontinuity designs (RDD) that generate robust estimates while minimising researcher discretion. Our results challenge the assumption of a uniform causal return to schooling for SRH. While FPE policies increased educational attainment and literacy in many cases - though not universally - they did not lead to improvements in women’s SRH outcomes on average. We find evidence of improvements only under specific conditions, particularly in settings characterised by very low baseline education and high prevailing fertility rates. We conclude that the abolition of school fees alone may be insufficient to generate equitable gains in schooling and to improve women’s reproductive health without complementary institutional conditions and supply-side investments.