Regional Educational Conditions and Childhood Stunting in East Nusa Tenggara: Spatial Evidence for Developmental Inequality

Abstract

Childhood stunting remains a persistent public health and developmental challenge in East Nusa Tenggara (NTT), Indonesia, where educational disparities may contribute to unequal nutritional outcomes across regions. This study examined global, local, and spatial relationships between educational conditions and childhood stunting across 22 districts and municipalities in NTT. District-level secondary data were analyzed using Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and Spatial Autoregressive (SAR) models. The educational indicators included mean years of schooling, educational resource availability, and community reading behavior. The OLS model explained 23.9% of the variation in stunting prevalence, with an adjusted (R2) of 11.2%, indicating relatively limited explanatory capacity when a single global relationship was assumed. GWR produced geographically varying local coefficient estimates and reduced residual spatial autocorrelation; however, Monte Carlo tests did not provide statistically significant evidence of coefficient non-stationarity. The SAR model further identified significant spatial dependence in stunting prevalence ρ = 0.578, p = 0.0005. After accounting for this dependence, mean years of schooling showed a significant negative association with stunting β = -3.899, p = 0.0056, whereas educational resource availability β =-1.392, p = 0.1028 and community reading behavior β = -1.071, p = 0.1747 were not statistically significant. These findings indicate that local coefficient estimates varied descriptively across NTT, although statistically significant coefficient non-stationarity was not detected. In contrast, stunting prevalence exhibited clear spatial dependence, while mean years of schooling was the only educational indicator significantly associated with stunting after spatial dependence was taken into account.