Experimental Design
The impact evaluation follows a cluster randomized controlled trial design with randomization at the school level. The final study sample consists of 120 primary schools distributed across two cohorts: Cohort 1 (2025–2026), comprising 42 schools, and Cohort 2 (2026–2027), comprising 78 schools.
Participating schools include both public and private-voucher institutions serving vulnerable student populations across multiple regions of Chile. Cohort 1 includes schools from the Metropolitan Region, O’Higgins, and Biobío regions. Cohort 2 includes schools from the Metropolitan Region, O’Higgins, Ñuble, and Biobío regions. For the purposes of randomization and analysis, schools in Ñuble are grouped together with those in the Biobío region.
Each cohort is followed for two consecutive academic years. In the first year, the evaluation includes students in grades 4 through 7; in the second year, the same schools are re-assessed, with the original student cohorts now in grades 5 through 8. Due to the staggered cohort structure, the second year of Cohort 1 overlaps calendar-wise with the first year of Cohort 2.
Schools are assigned to one of three groups: a high-support treatment arm (Intensivo), a moderate-support treatment arm (Semi-intensivo), or a control group. Both treatment arms receive the same core intervention, including standardized program materials, access to a digital platform, and common teacher training delivered prior to the first year of implementation and again between the first and second years. The two treatment modalities differ only in the intensity of ongoing implementation support provided during program delivery. Control schools continue with business-as-usual practices.
Randomization was conducted prior to baseline data collection using stratified random assignment, implemented separately by cohort. For Cohort 1, randomization was stratified by region, school dependency (public versus private voucher), and school enrollment size (above or below the mean enrollment within the Cohort 1 sample). For Cohort 2, randomization was stratified by region and school dependency only.