Evaluation of Arts-Integrated Social-Emotional Learning in Indian Government Schools

The intervention is an arts-integrated social-emotional learning (SEL) curriculum delivered through compulsory art classes in Punjab government schools. It is developed and implemented by Slam Out Loud (SOL), a non-profit with over ten years of experience working on arts-based education in Indian government schools, in partnership with Punjab's State Council of Education Research and Training (SCERT). The intervention bundles two components. First, a new SEL-integrated art textbook, aligned with India's National Education Policy (2020), is provided to every Grade 6 student in treatment schools. The textbook comprises five chapters delivered through approximately twenty lessons of forty-five minutes each. Second, teachers responsible for delivering the Grade 6 art curriculum in treatment schools participate in training organized by SCERT and SOL for three days, delivered directly by SOL trainers, covering all five chapters of the textbook, arts-based SEL facilitation skills, and classroom implementation. The intervention does not add instructional time, and only changes how the existing art period is used. Control schools continue with business as usual: the standard government art curriculum and textbook, with no additional teacher training, materials, or support from SOL.

2026-08-28

2027-01-15

Students' socio-emotional skills, measured through a battery of validated, age-appropriate scales collected at both baseline and endline: emotional awareness, emotion regulation, subjective wellbeing, curiosity, creativity (divergent thinking), and prosocial behaviour. Additionally, we will measure teacher-reported student behavior in the classroom using a modified version of the Social Behavior Questionnaire (SBQ).

We measure emotional awareness, a student's ability to identify and verbally share their own emotions, using a shortened Emotional Awareness Questionnaire that covers the differentiating-emotions and verbal-sharing subscales (Rieffe et al., 2008). We capture emotion regulation with the Emotion Regulation Questionnaire for Children and Adolescents (Gullone and Taffe, 2012). Subjective wellbeing, the child's recent emotional state and positive outlook over the prior two weeks, is measured with the Stirling Children's Well-being Scale (Liddle and Carter, 2015), dropping its built-in social-desirability items, which we capture separately. We measure curiosity using items adapted from Alan and Mumcu (2024), and creativity through an alternative-uses task in which students are given an everyday object and have two minutes to list as many uses for it as they can (Guilford, 1967). We measure prosocial behaviour through a stated willingness to help others and a hypothetical sharing decision in which the student allocates a gift of ten chocolates between themselves and a peer who received none (Falk et al., 2018; Kosse et al., 2020). We additionally measure students' social behaviour as rated by their teacher, using an adapted short form of the Social Behavior Questionnaire (SBQ; Tremblay et al., 1991; Collet et al., 2023), covering the prosocial behaviour, disruptive behaviour, and hyperactivity-inattention subscales; the teacher responsible for the Grade 6 art period rates a random subset of students per class, drawn from the baseline enrolment roster, and this measure is collected at endline only. Each scale is scored following its source instrument, with the indicated items reverse coded so that higher values denote more of the outcome being measured, and then standardized so that treatment effects are expressed in control-group standard deviations.

Learning and school outcomes: academic performance, measured using students' final-year Grade 6 examination scores obtained from administrative records; attendance, measured from Grade 6 attendance records; and classroom behaviour, captured through structured classroom observation. Teacher outcomes: teachers' teaching and learning beliefs, self-reported teaching practices, their own emotion regulation, art self-efficacy, and perceived student engagement.

The teacher who delivers the Grade 6 art period is surveyed in all 200 study schools at both baseline and endline. We examine teachers' teaching and learning beliefs, covering growth mindset and openness to new pedagogy; their self-reported teaching practices along a student-centred versus traditional dimension; their own emotion regulation, using the adult Emotion Regulation Questionnaire (Gross and John, 2003); their art self-efficacy and enjoyment of teaching the art period; and their perceived student engagement. These outcomes let us test whether the training shifts how teachers think about learning and how they teach, a plausible channel for effects on students. The teacher who handles the art period usually teaches a regular subject too, often to the same Grade 6 students; using a baseline item that records whether the art teacher also teaches their main subject to the same class, we test for spillovers to students' grades in that subject.

We use a school-level cluster randomized controlled trial. 200 schools were randomized to treatment and control in equal proportion, stratified by district. Schools are randomly assigned to treatment or control before the baseline survey. Our primary analyses follow an intention-to-treat specification based on randomized assignment: we compare outcomes for students in treatment and control schools at endline, with district fixed effects and school-level baseline controls, and an ANCOVA specification that additionally controls for the baseline value of the outcome measured at the student level. Standard errors are clustered at the school level, the unit of randomization.

Not available

Randomization done via STATA by the research team, using school-level administrative data, stratified by district.

The school is the unit of randomization because the intervention is delivered at the school level, through the teacher conducting Grade 6 art classes. There is a single level of randomization.

Yes

200 government schools

Approximately 5,400 Grade 6 students across 200 schools, and approximately 200 teachers.

100 schools treatment, 100 schools control

Under conservative assumptions (ICC = 0.05, baseline-endline correlation of the outcome indices of 0.30), we can detect a minimum effect size of approximately 0.09 standard deviations with 80% power at the 5% significance level. Under more optimistic assumptions (ICC = 0.01, baseline-endline correlation of 0.50), the minimum detectable effect decreases to approximately 0.06 standard deviations. Under the assumption of 85% take-up, our MDE increases to approximately 0.10 standard deviations.