Experimental Design
Selection of Locations
Amhara and Oromia: Five woredas in South Gondar, Amhara (Ebenat, Lay Gayint, Libo Kemkem, Simada, and Tach Gayint) and five woredas in East Hararghe, Oromia (Fadis, Babile, Jarso, Haramaya and Gursum) were purposely sampled for inclusion in the study.
These two zones were selected based on food insecurity status and high rates of child marriage (especially among the 10-14 age bracket), as a proxy of conservative gender norms and relatively less programming on child marriage. Within these 10 woredas, kebeles were characterized into three groups: (i) unsafe for data collection and programming, (ii) marginalized (lack of programming, isolated from key services and road/transport infrastructure) and (iii) less-marginalized (in terms of access to services and to the main woreda town). Locations identified by local officials as characterized by high security concerns were excluded from consideration.
Afar: In Afar, all five woredas (Dalifage, Dewe, Hadelela, Semurobi, and Telalak) in Zone 5 were selected for inclusion in the study. Zone 5 was chosen because of its relative marginalization within the Afar region - in terms of poor infrastructure (including distance from main roads to Semera, Tigray and Djoubti), high proportion of households who migrate seasonally, weak political representation (due to distance from Semeara the regional capital and less politically influential clans) and very low education rates, limited economic opportunities (e.g. distance from the salt mines and trading routes with Tigray region which other Zones enjoy) as well as ethnic tensions.
Randomization
Amhara and Oromia: Prior to the start of baseline data collection, the 155 kebeles were block randomized (by woreda, and kebele marginalization status) into control and four hypothetical treatment arms (the specific arms were not known at the time of this initial randomization). Within each of these blocks, the first kebele listed was the control, and then the four treatment arms after that (Her Spaces, Act with Her, Act With Her (curriculum only), and Act With Her + Assets). In almost all kebeles there was one extra kebele that was assigned control.
Afar: These 20 kebeles were organized into 10 blocks (by woreda and kebele marginalization status). Each block contained two kebeles. At the time of baseline data collection, there was no indication that interventions would take place in Afar. As soon as we learned that AWH would operate in Afar, we randomly allocated one of each pair to treatment and one to control. Within each of these blocks, the first kebele listed was the control, the next as Act With Her, then control again, and so on.
Adolescent inclusion and exclusion criteria: All sites
The listing exercise in each location sought to identify all adolescent boys and girls in the household aged 10-12 years old (at the time of listing). If the household had more than one individual in this age range, one adolescent was randomly selected for study eligibility.
It is worth noting that GAGE baseline data collection took place in 2017 prior to the trial start date as part of the broader GAGE longitudinal data collection. While there was an aspiration for an eventual RCT, this original data collection was done in the absence of a planned trial. This trial will utilize these sites and the 2017 data as its baseline.