GEOGRAPHICALLY WEIGHTED REGRESSION WITH DIFFERENT KERNELS: APPLICATION TO MODEL POVERTY

Abstract

Poverty is still a significant problem in Indonesian development. The poverty alleviation programs implemented have yet to pay attention to spatial aspects, so the policies are often not on target. This study aims to reveal the spatially varying relationships between the poverty level and its factors at the regional scale and compare three fixed kernels as a weighting matrix for GWR. The method used is geographically weighted regression (GWR) with poverty data for 2021. The study results show spatial autocorrelation and is grouped in 29 regencies/cities. The expenditure per Capita, life expectancy, percentage of houses and households with proper drinking water, open unemployment rate, labor force participation rate, and GDP at constant prices show different effects in each region. The results strengthen the argument that spatial aspects cannot be ignored in regional development, especially poverty alleviation. Therefore, area-based poverty alleviation can be used as a basis for determining/determining policies so that they can be more targeted.