The state of evidence for social and emotional learning: A contemporary meta-analysis of universal school-based SEL interventions

424 studies, 575,361 students, 53 countries, ages 5-17, published 2008-2020, covering 252 distinct interventions

School students aged 5-17, international.

A Registered Report, which is the strongest design feature available to a meta-analysis, using correlated-hierarchical-effects models with robust variance estimation. 237 of 424 studies were randomised. Nearly 73% used education-as-usual or waitlist controls. Only 44.6% reported any implementation data and of those only 21% reached moderate or high fidelity. Only 55% of studies were conducted in the intervention's country of origin. Independence is developer-involved at field level - Durlak is a co-author of the 2011 meta-analysis this one revises, and the lead author directs a SEL research and advocacy centre. The tone is nonetheless self-critical; this is the field auditing itself.

Overall g = 0.194 (0.166 to 0.221) across 1,865 outcomes, with I-squared of 93.7%. By domain - school climate 0.293, civic outcomes 0.255, peer relationships 0.222, SEL SKILLS 0.219 (against 0.57 in Durlak), attitudes 0.209, prosocial 0.178, externalising 0.162, emotional distress 0.140, school functioning 0.122. ACADEMIC ACHIEVEMENT g = 0.109 (0.037 to 0.182), roughly 60% below Durlak's 0.27. Null results on discipline (0.183, CI -0.237 to 0.602, p = .328), family relationships and physical health. Follow-up, available for 11% of studies at a mean of 53 weeks, gave 0.167 overall with 6 of 12 domains still significant. PUBLICATION BIAS IS SIGNIFICANT - Egger t = 3.59 overall (p < .001), with published studies at 0.220 against unpublished at 0.164. On measurement, 53.5% of outcomes are child-reported and 58.7% are rating scales, checklists or questionnaires.

Medium - 44% of the pool is quasi-experimental.

Partial and attenuating. It reproduces a positive overall signal at about two thirds the size on skills and 40% the size on achievement, and its own publication-bias tests are significant.