Modeling the Percentage of Poor Population in Java Island using Geographically Weighted Regression Approach

Abstract

Poverty is a multidimensional problem faced by all countries in the world. Poverty is the inability of individual or group to meet their basic needs in terms of expenditure. In poverty problem, there is a tendency that the poor will group in locations with certain characteristics. This spatial clustering indicates spatial diversity that making global regression analysis inappropriate for application. Therefore, the purpose of this research is to model the percentage of poor population in 119 districts on Java Island in 2021 using the Geographically Weighted Regression (GWR) method. The analysis results state that the GWR model with Kernel Fixed Bisquare provides superior results compared to the global regression model and able to overcome spatial heterogeneity problem. The model is able to provide a fairly high coefficient of determination, which is 70,73 percent. The GWR model identifies ten groups of districts based on the significance of the independent variables, with the majority of them (61 districts) having a significant RLS variable. This indicates that education is an important aspect that needs to be considered by local governments to alleviate poverty.