Changing Gang Paths at Adolescence: Long-Term Effects of a Behavioral Intervention

The intervention is a school-based after-school program (ASP) implemented by Glasswing International in five public schools located in highly violent communities in El Salvador. The program was delivered over 20 weeks between April and October 2016, targeting students in grades 4 through 9. Clubs met twice per week for approximately 90 minutes, immediately after the school day, on school premises.

Each session had two components. The first focused on behavioral and socioemotional skills development, uniform across all clubs. Drawing on psychology-based principles related to emotion and behavior regulation, it aimed to strengthen self-regulation and decision-making through activities emphasizing impulse control, emotion regulation, perspective-taking, and planning, delivered via experiential learning, role-play, and guided reflection.

The second component consisted of structured extracurricular activities designed to foster engagement and sustained participation. These varied by club type across four categories: leadership, art and culture, sports, and science. This component provided a supervised, socially interactive environment in which students could practice the behavioral skills introduced in the first part of each session.

Sessions were led by volunteer tutors who were non-specialists but received standardized training and detailed session manuals. This delivery model was designed to facilitate scalability in settings with limited access to mental health professionals. The mean club size was 13 students, and average take-up was 57%.

2016-05-02

2016-09-30

The primary outcomes fall into the following domains:

Educational outcomes are measured using administrative records from the Ministry of Education (MINED) through the national student information system (SIGES) for the period 2017–2024. These include: (i) completion of lower secondary school (9th grade), (ii) dropout before completing high school, (iii) participation in the national standardized exam (PAES), and (iv) PAES scores.

Criminal outcomes are measured using administrative records from two sources. Gang membership is measured using intelligence records from the National Civil Police (PNC), which compiles gang recruitment and profiling data through field informants. Criminal offenses and arrests are measured using incarceration records from the General Directorate of Prisons (DGCP). Specific outcomes include: (i) gang membership, (ii) whether the individual committed a crime, and (iii) whether the individual has been arrested.

Labor market outcomes, conditional on data access, include formal employment status and wages, measured using administrative records from the Salvadoran Social Security Institute (ISSS). These outcomes will be analyzed if the data is made available by the institution.

The experimental design allows estimation of both the average long-term impacts of the after-school program and the role of peer group composition in shaping these impacts.

From approximately 2,420 eligible students enrolled in grades 4 through 9 across the five schools, 1,056 students applied and were enrolled in the 2016 studies (Dinarte-Diaz, 2024 and Dinarte-Diaz and Egana-DelSol, 2024). Randomization proceeded in two stages within school-by-educational-level strata.

First, enrolled students were randomly assigned at the individual level to either a control group (25 percent) or a treatment group offered participation in the ASP (75 percent). Students assigned to the control group did not participate in ASP activities and left school premises at the end of the regular school day.

Second, students assigned to treatment were randomly allocated to clubs that differed in peer composition based on a predicted baseline propensity-for-violence index. This index was constructed using external administrative and survey data on youth violence in El Salvador and combines individual, household, and school characteristics associated with exposure to violence. Students were assigned to either (i) heterogeneous clubs, in which peers differed widely in baseline propensity for violence, or (ii) homogeneous clubs, in which peers had similar baseline risk levels. Within homogeneous clubs, students were further grouped into high-risk and low-risk clubs based on whether their index lay above or below the median within each stratum. Neither students nor tutors were informed of the rationale behind group assignments. Further details are presented in Dinarte-Diaz (2024).

Not available

Randomization done in office

Individual (student-level)

No

NA

1056 students

The experimental sample consists of 1,056 students. Within school-by-educational-level strata, students were randomly assigned to three arms:
Control group: 258 students (25 percent of the sample), who did not participate in the ASP.
Heterogeneous treatment group: 263 students, assigned to clubs composed of peers with diverse baseline propensities for violence.
Homogeneous treatment group: 535 students, assigned to clubs composed of peers with similar baseline risk levels. Within this arm, students were further grouped into high-risk and low-risk clubs based on whether their predicted violence index was above or below the median within each stratum.