China’s Double Reduction Policy and Adolescent Depressive Symptoms: Longitudinal Evidence on Afterschool Activity Reallocation

Background China’s Double Reduction policy was implemented starting in the autumn 2021 school term and reorganized homework, school-provided after-school services, and off-campus tutoring for compulsory education students. Evidence from a long-standing household panel with repeated measurements of depressive symptoms and a contemporaneous upper-secondary comparison remains limited. Upper secondary students were outside the homework and school service provisions of compulsory education, although related governance was extended to off-campus subject tutoring for upper secondary students. Methods: Data from the 2018, 2020, and 2022 China Family Panel Studies were analyzed. The main analytical sample consisted of 1,005 students observed in both 2020 and 2022. Depressive symptoms were measured with the eight-item Center for Epidemiologic Studies Depression Scale (CES-D8). Predicted grades 7–9 in autumn 2021 defined the group with more intensive coverage, while predicted grades 10–11 formed a comparison group with lower policy intensity rather than an untreated group. Individual fixed-effects difference-in-differences models were adjusted for the current inverse hyperbolic sine of per capita household income and household size. The available prepolicy diagnostic, adjacent-grade comparison, alternative outcome models, principal component analysis (PCA) activity indices, fixed-effects pathway models, subgroup models, and pooled interaction tests were examined. Results: More intensive coverage was associated with a 0.925-point larger decline in CES-D8 (95% CI − 1.450 to − 0.401; p < 0.001), approximately 0.26 baseline standard deviations. The adjusted 2018 interaction was − 0.006 (p = 0.982), whereas the 2022 interaction was − 0.880 (p = 0.003). Non-academic organized training declined, restorative activity and weekly exercise increased, and sleep duration increased, though the increase was not statistically significant. Academic burden, policy-targeted training, and digital activity showed no detectable differential change. A favorable sleep pathway and an offsetting talent-training pathway were both marginally significant. Within-group estimates were significant among males and middle-income students, but the pooled interaction tests were not significant. Conclusions: More intensive coverage was associated with modestly more favorable depressive symptom trajectories and selective activity reallocation. The comparison is conservative because upper secondary tutoring was also subject to related governance. Developmental change, pandemic conditions, and mediator outcome confounding preclude definitive causal claims.