Building Teacher Capacity or Outsourcing Pedagogy? Experimental Evidence from Science Education Reform

We evaluate two delivery models for curiosity-based science education in government middle schools in rural Uttar Pradesh, India. The intervention introduces a ten-session, hands-on science curriculum—developed in partnership with the Agastya International Foundation—that promotes inquiry, experimentation, and reflection using low-cost, locally available materials.

In the Teacher Training (T1) arm, regular government science teachers receive multi-day in-person workshops and ongoing coaching from the implementing partner Ecoprism. The training focuses on integrating curiosity-based pedagogy into regular classroom instruction, modeling active-learning techniques, and supporting teachers through moderated peer groups.

In the External Instructor (T2) arm, trained Ecoprism facilitators directly deliver the same ten lessons to students during school hours while regular teachers continue standard instruction. This model allows high-fidelity delivery independent of teacher capacity constraints.

Both interventions use identical curricular content and implementation materials, differing only in who delivers the instruction. The design isolates two common policy approaches in education reform—building internal teacher capacity versus outsourcing delivery to specialized external agents—and measures their relative effects on student curiosity, critical thinking, growth mindset, and learning outcomes.

2025-08-22

2026-01-15

The primary outcomes are measured at the student level:

(i) Curiosity Index: Composite built from items adapted from Alan & Mumcu (2024) and the curiosity subscale of QATCT (Manassero-Mas et al., 2022). 5-point Likert items; aggregated via Anderson (2008).
(ii) Critical Thinking Disposition Index: Items adapted from the Critical Thinking Disposition Scale (Sosu, 2013) capturing openness to new information, reflection, intellectual humility. 5-point Likert; Anderson (2008) aggregation.
(iii) Cognitive Skills & Academic Performance: Curriculum-aligned Science Test Score (4 MCQs; grade-appropriate items drawn from district exams/textbook content), ASER Maths Score (3 items: two subtraction, one adapted multiplication), Literacy Score (PASEC-style reading comprehension items in Hindi), Raven’s Progressive Matrices Score (5 colored items), Administrative final science/overall grades (where available).

To address multiple hypothesis testing, we will control the false discovery rate within outcome families and report corresponding q-values (Benjamini and Heller, 2007).

Secondary outcomes capture broader psychosocial and behavioral mechanisms that may mediate learning impacts on students. These include:
(i) Study Preferences and Aspirations: Tracks students’ interest in science, intended pursuit of STEM subjects, and collaborative study habits to assess engagement and peer learning.
(ii) Growth Mindset: Measures students’ beliefs about effort and the malleability of intelligence, capturing adaptability and perseverance in learning.
(iii) Student-Reported Classroom Activity: Assesses classroom participation, questioning, and teacher feedback to gauge shifts toward interactive, inquiry-based instruction.
(iv) Self-Reported Time Use: Records daily time spent on study, tutoring, chores, and leisure to detect changes in study effort and subject focus.

Additionally, we will collect outcomes for teacher, including:
(i) Curiosity: Assesses teachers’ intrinsic motivation and interest in learning, using a scale parallel to the student curiosity measure for comparability.
(ii) Cognitive Skills: Tests teachers’ mastery of core science concepts through a short assessment aligned with the student curriculum.
(iii) Teaching Philosophy: Captures teachers’ pedagogical beliefs, self-efficacy, and openness to adopting inquiry-based, student-centered instruction.
(iv) Self-Reported Time Use: Records how teachers allocate time across instructional, administrative, and collaborative tasks inside and outside the classroom.
(v) Knowledge of Students: Measures teachers’ familiarity with their students by comparing teacher estimates of student traits to actual student responses.

We will also collect externally observed measures of classroom practice and instructional quality.

The study uses a cluster-randomized controlled trial with randomization at the school level. Randomization is stratified by administrative block. A total of 150 government middle schools in Fatehpur district, Uttar Pradesh, are randomly assigned to one of three groups:

Treatment 1 (Teacher Training) – science teachers receive in-person workshops and ongoing coaching to integrate curiosity-based, inquiry-driven pedagogy.

Treatment 2 (External Instructor) – trained facilitators from the implementing partner deliver the same ten-session curiosity curriculum directly to students.

Control – business-as-usual classroom instruction with no exposure to the new curriculum.

Not available

Randomization done in office by a computer

School

Yes

150 schools

7185 students

50 schools control, 50 schools in treatment 1 (teacher training), 50 schools in treatment 2 (external instructor)

Minimum detectable effect size (comparison between pooled treatment and control) is 0.146 sd for curiosity index, 0.142 sd for critical thinking index, 0.12 sd for growth mindset, 0.148 sd for science test score.