Personalized Education Mentor Program: Main Randomized Controlled Trial in Meizhou Primary Schools
How to cite
Aucejo, Esteban; Chau, Chunchiu; Qiu, Xincheng. 2026. "Personalized Education Mentor Program: Main Randomized Controlled Trial in Meizhou Primary Schools." AEA RCT Registry. AEARCTR-0018426. https://doi.org/10.1257/rct.18426-1.1
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
This registration covers the main Phase II randomized controlled trial of a school-based mentorship intervention in Meizhou, China, and is distinct from an earlier pilot conducted in a different location and registered separately. The study population consists of primary-school students in participating schools in Meizhou, covering grades 1 to 6 and approximately 3,147 students in total. The intervention assigns trained mentors to a subset of students during the study period. Randomization occurs in two stages. First, classrooms are randomized within school-by-grade strata into treatment-cluster classrooms or pure-control classrooms. Second, within treatment-cluster classrooms, students are randomized individually into mentorship treatment or within-class control. The primary estimand is the direct mentorship effect, comparing treated students to untreated students within the same treated classroom. A secondary estimand is the spillover effect, comparing untreated students in treated classrooms to students in pure-control classrooms. Primary outcomes include academic achievement based on school administrative performance measures (using administrative scores in Chinese, Math, and English, where available). Secondary outcomes are socio-emotional/non-cognitive indices constructed from the student survey measures (grit, mental health, confidence, help-seeking, self-academic perception, independent thinking, anxiety, social skills, and expectations). The planned sample is approximately 67 classrooms and about 3,147 students.
Experimental design
The study is a two-stage randomized controlled trial conducted in three primary schools in Meizhou, China. Classrooms are first randomized within school-by-grade strata into one of two classroom environments: treatment-cluster classrooms or pure-control classrooms. The classroom allocation is designed to be as close as possible to a 2:1 ratio of treatment-cluster classrooms to pure-control classrooms, subject to the number of classrooms available in each stratum and integer constraints. Within treatment-cluster classrooms, students are then randomized individually into mentorship treatment or within-class control in an approximately 50:50 ratio. Student-level randomization is stratified by gender and, when the necessary baseline roster is available before assignment, by a coarse baseline achievement indicator. This design yields three final groups: treated students in treated classrooms, untreated students in treated classrooms, and students in pure-control classrooms. The main analysis will focus on the direct mentorship effect, defined as the comparison between treated students and untreated students within treated classrooms. A secondary analysis will examine spillover effects by comparing untreated students in treated classrooms to students in pure-control classrooms. Analyses will include the school-by-grade strata used in randomization, baseline controls where available, and standard errors clustered at the classroom level.
Mentor Randomization: Mentors are also randomized based on mentors' availability. For example, if at a given school on Tuesday, we have 168 students in Grade 1. Based on our 1:6 ratio, we have approximately 28 mentors. These 28 mentors are randomly assigned to these 168 students. We have established four dedicated mentor groups (assigned by day: Tuesday through Friday). Each group is responsible for all sessions on their assigned day to ensure consistency in instruction and logistics. Throughout all sessions, we maintain a strict 1:6 mentor-to-student ratio.
Design details
Cited by
No citation data available for this registration yet.