The Effect Of School-Age Population Decline On Elementary School Closures And Consolidations In South Korea: Evidence From A Municipality-Year Panel Using Double Machine Learning And Causal Forests

Abstract

South Korea’s exceptionally low fertility and uneven population redistribution are shrinking elementary school-age cohorts, raising concerns about the sustainability of local school networks. This study examines whether municipality-level declines in the population aged 6–11 are associated with elementary school closures and consolidations. We compile a municipality-year panel covering all 229 Si/Gun/Gu municipalities for 2019–2024 (1,374 observations) and link demographic measures to public elementary school structure and annual closure/consolidation actions. We estimate two-way fixed-effects models and apply double machine learning with cross-fitting to flexibly adjust for observed covariates, and we explore heterogeneity using interaction models and forest-based approaches. Consolidation actions are rare but spatially concentrated: 7.35% of municipality-years record at least one action. Average within-municipality effects of short-run cohort change are small and statistically imprecise. However, effects are heterogeneous: municipalities with structurally small baseline school systems exhibit greater sensitivity to cohort decline, and a forest-based severe-decline contrast reveals wide dispersion in conditional effects. The findings suggest that demographic decline acts as a structural pressure that triggers consolidation primarily in vulnerable local systems. Policy responses should prioritize place-based early-warning screening and mitigation measures in high-sensitivity municipalities rather than uniform consolidation rules.