Prevention of GBV in Kenya

The intervention is a school-based sexual violence prevention and empowerment curriculum delivered by trained public school teachers in Kenya, which focuses on skills to prevent and respond to sexual harassment/assault and promote safer school environments.

2025-03-01

2025-10-31

1) Incidences of sexual assault. 2) School attendance. 3) Teen pregnancy and teen marriage (conditional on feasibility of 2026 follow-up tracking).

Sexual assault is measured using girls’ self-reports of forced sex (binary and count, based on whether it occurred and how many times in the past year). Attendance is measured using school administrative attendance records and cross-validated using student self-reports (e.g., days missed and prolonged absence; where feasible, dropout). Teen pregnancy and teen marriage are measured (if 2026 follow-up tracking is feasible) via school roll-call tracking of girls who dropped out due to pregnancy or marriage since the intervention, with possible verification on a subsample of dropouts if logistics permit.

Girls: self-defense skills; awareness of sexual consent; personal initiative (self-efficacy, perseverance/resilience/problem-solving, and agency); perceived safety at school; perceived safety commuting to/from school; rapport and interactions with teachers; relationships with opposite-sex peers (in co-ed schools).
Boys: beliefs about sexual assault, consent, and consequences/deterrence; bystander intervention behaviors; masculinity and help-seeking behaviors; relationships with opposite-sex peers (in co-ed schools).

Secondary outcomes are measured using student questionnaires at baseline/endline (and teacher questionnaires during teacher trainings). Where multiple survey items capture a construct, outcomes will be aggregated into indices (standardized to mean 0 and SD 1). For some domains we use PCA (first principal component) and for others we use summary indices (including inverse-covariance weighted indices following Anderson (2008) when combining conceptually related components). Examples: girls’ self-defense index aggregates knowledge and reported verbal/physical self-defense; consent-awareness is the first PC of “ability to say no” items; PI sub-indices (self-efficacy, perseverance/resilience/problem-solving, agency) are constructed via PCA and then combined depending on correlations; boys’ beliefs and masculinity/help-seeking are constructed via PCA; intervention behavior is constructed from exposure-to-situation and reported intervention items.

Cluster randomized controlled trial at the school level in Migori County, Kenya, comparing schools assigned to receive the intervention curriculum versus control schools.

Not available

Randomization was conducted by computer using a stratified randomization algorithm.

School (cluster) level.

Yes

214 schools.

Approximately 8,560 students (about 40 students per school).

107 treatment schools and 107 control schools.

Using 214 schools (107 treatment, 107 control) with approximately 20 surveyed students per school per gender and clustering at the school level, the minimum detectable effect (MDE) is approximately 0.09 SD using the baseline ICC estimate (ICC ≈ 0.009). Under more conservative ICC assumptions, MDEs are approximately 0.12 SD (ICC = 0.05) and 0.15 SD (ICC = 0.10), at 5% significance and 80% power.