Published on Development Impact

Can an after-school program reduce adolescent violence at school in developing countries?

For the past few years, I have worked to understand the causes and consequences of crime and violence, how these issues affect human capital accumulation, and what interventions can address these behaviors. When Jen Doleac interviewed me a couple of years ago on the podcast Probable Causation, she asked how I came to focus on this field. I told her my interest stemmed from witnessing firsthand how adolescents in my native El Salvador encountered danger on a daily basis, and how this hindered their human capital accumulation. My own personal observations have been extensively confirmed by an array of evidence. Once out of school and on the streets, these vulnerable youths are likely to be recruited by gangs, thereby perpetuating the vicious cycle of violence and death.

One specific approach to address exposure to or participation in violent behaviors among young people has been after-school programs (ASPs) for adolescents. These interventions are implemented in school facilities after the school day and usually include activities related to learning and extracurricular activities (sports, the arts, leadership, etc.). Overall, evidence on the effects of ASPs mostly features data from developed countries (such as Durlak et al. 2011). Evidence from developing countries is still scarce and focuses on improving academic performance rather than altering behavior and social-emotional skills, which are arguably equally (or more) important to students in highly violent communities.

Why ASPs? They provide at least three important services. First, ASPs can protect children by keeping them occupied at a time when they might otherwise be unsupervised and exposed to external risks, thereby preventing victimization and delinquent behavior. Second, when these programs include a specific curriculum to foster social-emotional skills and help students control impulsivity, they offer an alternative source of social-emotional learning and development. Third, the adults who lead and support such programs serve as role models for participants. Other benefits could include positive interaction with peers and the acquisition of new skills through extracurricular activities.

The first ASP I  evaluated was implemented in El Salvador in 2016 through a partnership with the NGO Glasswing International, the largest organization to offer school-based ASPs in Central America. We designed an experiment to examine several questions. First, how do ASPs affect students’ violent behaviors, attitudes toward school and learning, and academic performance? Second, is emotional self-regulation a mechanism (Dinarte-Diaz and Egana-DelSol, 2022)? Third, using the same intervention and experimental design, how can group composition maximize the effectiveness of an ASP on the same set of outcomes (Dinarte-Diaz, 2020)? That is, does participation in an ASP with a particular composition of peers (in terms of propensity for violence) cause greater impacts?

To simultaneously respond these two questions, I designed a tracking-by-violence experiment (similar to Duflo, Dupas and Kremer, 2011). First, I randomly assigned students to three groups: (i) participation in the ASP with a group of peers with different propensities for violence (heterogeneous groups), (ii) participation in the ASP with a group of peers with similar propensities for violence (homogeneous groups), and (iii) a comparison group. Then, I ranked students within group (ii); those with a propensity for violence were assigned to a high-violence homogeneous group and the others to a low-violence homogeneous group. To answer the first question, I compared students in groups (i) and (ii) with group (iii). To answer the second question, I compared group (i) to group (ii).

To conduct this experimental design, I needed a measure of violence. Unfortunately, as researchers working in violent settings have noted, it is challenging to collect high-quality data in these environments. In El Salvador, one concern about collecting data on violence at baseline was that we could not guarantee the confidentiality of this personal information (e.g., local authorities or gangs might force researchers or the NGO to reveal information that identified each child), risking not only the intervention but also—most importantly—the children’s safety. To avoid such issues, I estimated a predictive model of violence from existing data and participants’ sociodemographic characteristics.

In Dinarte-Diaz and Egana-DelSol, (2022), my coauthor and I find that the intervention reduced school-based misbehaviors and other violent behaviors by 0.22 SD and 0.14 SD, respectively. The program also reduced the number of absences (which can be seen as a protection factor) by 28 percent. Although the program did not explicitly target academic learning, a better classroom environment and fewer absences might have indirectly affected students’ grades as well. Indeed, we find that ASP participants improved their academic grades by 0.15 SD. Automaticity—that is, the likelihood to respond to situations automatically and without deliberation—and impulse control are the mechanisms we explore here. Using recordings from students’ brains measured through electroencephalograms and a lab-in-the-field setup, we show that ASP participants became calmer and more emotionally and behaviorally composed compared to the control group.

In terms of peer effects, in Dinarte-Diaz, (2020), I show the ASP’s effects on the main outcomes were larger when students participated with a more diverse set of peers. Interestingly, this was true not only for less violent but also for more violent peers. These results align with the hypothesis that students in heterogeneous groups are exposed both to good behaviors they should emulate and negative ones they should reject. These interactions were less available to students in the homogeneous group. In line with these main results, I also find that students assigned to homogeneous groups had stress levels greater than those of students placed in a heterogeneous setting.

Overall, these results have promising implications for public policy discussions regarding interventions designed to reduce violence within schools and, indirectly, to improve academic outcomes. This evidence is relevant to educational systems in developing countries that lack the necessary resources to provide quality education, but can potentially offer low-cost after school programs at the school.

There are, of course, other questions relevant to the replication or scale-up of ASPs. For example, what is the most important component: protection or social-emotional learning? What role do tutors play and how do they mediate the effects? Does curriculum type matter? What type of curriculum is more relevant: a light-touch, psychology-based intervention or a highly intensive one? In Dinarte-Diaz, Egana-DelSol and Martinez (2022), we discuss some of these questions in the context of Central America. I will provide further insights about the results in future blog posts.

This work was enriched by our lab-in-the-field with novel tools such as electroencephalograms, task-based games on tablets, and others. I will delve more into how we developed and implemented these components in future blogs. So stay tuned!


Lelys Dinarte-Diaz

Research economist in the Human Development Team of the World Bank's Development Research Group

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000