Stratified Cluster Randomization Design
RCTJournal Article
Assigning educational interventions at the individual student level within classrooms creates severe administrative disruption and risks cross-treatment spillovers between peers.
Picture this
Grouping entire schools into balanced clusters based on baseline test scores, language, and gender before tossing a coin to assign treatment functions like sorting sports teams into equal divisions before randomly selecting which half receives specialized coaching equipment.
What the evidence says
Stratification achieved complete baseline balance across treatment and comparison groups, keeping all initial pretest score differences below 0.10 standard deviations.
- Who was studied
- N = 12,855 students in Year 1 and N = 21,936 students in Year 2 across primary government schools in Vadodara and Mumbai, India.
- How
- Cluster-randomized controlled trial with pre-assignment stratification by baseline pretest scores, language of instruction, and gender ratio.
What to do
Group schools or clusters by baseline academic performance, language, and gender before executing random assignment to treatment arms in educational trials.
From the source
"To ensure a balanced sample, assignment was stratified by language, pretest score, and gender."
Remedying Education: Evidence from Two Randomized Experiments in India