GEOGRAPHICALLY WEIGHTED PANEL REGRESSION MODELING OF POVERTY RATES IN TROPICAL RAINFOREST AREAS OF KALIMANTAN

Abstract

When applied to spatial panel data, the Geographically Weighted Panel Regression (GWPR) model is a localized version of the linear regression model. The Fixed Effect Model (FEM) inside estimator is used as a global model in this investigation. The purpose of this research is to obtain a GWPR model and identify the variables that affect the proportion of the impoverished in 56 districts and cities located in Kalimantan's humid tropical forest region between 2019 and 2022. The Weighted Least Square (WLS) approach, which provides geographic weighting in addition to the Least Square method, is used for estimating the parameters of the GWPR model. The optimal weighting function chosen from the adaptive bisquare, adaptive tricube, and adaptive gaussian weightings is the spatial weighting function used in the GWPR model estimate in this work. For determining the ideal bandwidth, the Cross Validation (CV) criterion is applied. According to the study's findings, the optimal weighting function is adaptive gaussian, which yields the best GWPR model with a CV of 8.8740 at the lowest. The GWPR model parameters were tested, and the results showed that both local and global influences affect the percentage of the population living in poverty. The gross domestic product (GDP), the open unemployment rate, the average length of education, the number of workers, and life expectancy are local factors that affect the percentage of the poor; on the other hand, the number of workers is a global factor that affects the percentage of the poor.