<![CDATA[The Geographically Weighted Panel Regression (GWPR) model combines panel and spatial data to capture geographic heterogeneity by allowing variable effects to differ by location. This quantitative study examines poverty rates in Central Java Province from 2022 to 2024 using GWPR, analyzing secondary data from 35 districts/cities provided by the Central Java Provincial Statistics Agency (BPS). Independent variables include Gross Regional Domestic Product (GRDP), labor force participation, minimum wage, literacy, school participation, sanitation, and clean water access. This study examines the spatial–temporal determinants of poverty in Central Java Province using a spatial-panel modeling approach. Panel regression analysis was first conducted, and the Chow, Hausman, and Lagrange Multiplier tests identified the Random Effect Model (REM) as the most appropriate global specification. However, evidence of spatial heterogeneity suggested that global parameters could not adequately capture interregional differences. To address this limitation, Geographically Weighted Panel Regression (GWPR) was employed to simultaneously model spatial and temporal variation. Estimation was performed using Weighted Least Squares with a bisquare kernel, and the optimal bandwidth was selected through Cross-Validation, yielding a minimum CV value of 0.0038 with a bandwidth of 9. The GWPR model achieved a markedly higher R² (0.9996) than REM (0.5628), indicating superior capacity to represent localized structural variation. The range of Local R² values (0.242–0.898) demonstrates heterogeneous model fit and reduces concerns of overfitting, with bandwidth selection functioning as a nonparametric regularization mechanism. These findings highlight the importance of spatially adaptive poverty policies tailored to district-specific socioeconomic conditions in Central Java Province, Indonesia. ]]>
