IMPLEMENTATION OF GEOGRAPHICALLY WEIGHTED PANEL REGRESSION WITH GAUSSIAN KERNEL WEIGHTING FUNCTION IN THE OPEN UNEMPLOYMENT RATE MODEL

Abstract

This study analyzes the factors influencing the Open Unemployment Rate in Kalimantan using the Geographically Weighted Panel Regression (GWPR) model with Gaussian kernel weighting functions. The GWPR model, a local panel regression approach for spatial data, is compared with the global Fixed Effect Model (FEM). Spatial weighting for parameter estimation employs Fixed Gaussian and Adaptive Gaussian kernels, with the optimum bandwidth determined through Cross Validation (CV), resulting in a minimum CV value of 25.536 for the Adaptive Gaussian Kernel. Local factors identified as influencing the Open Unemployment Rate include the Labor Force Participation Rate ( ), Expected Years of Schooling ( ), Average Years of Schooling ( ), Total Population ( ), Number of Poor People ( ), and the Growth Rate of Gross Regional Domestic Product at Constant Prices ( ). The results underscore the importance of spatial heterogeneity in understanding regional unemployment dynamics, as local variations in these factors significantly affect unemployment rates. Moreover, the GWPR model exhibits a notable improvement in predictive accuracy and goodness of fit compared to the global panel regression model, achieving a coefficient of determination  of 77.96% and a Root Mean Square Error (RMSE) of 0.2726. These findings highlight the GWPR model's potential in regional economic studies and policymaking, offering precise insights into local determinants of unemployment and facilitating the development of targeted and effective interventions.