Modeling Poverty Data in Indonesia with Spatial Hierarchy Structure Using HLM, GWR, and HGWR Methods

Abstract

Poverty causes the majority of the Indonesian population to face challenges in fulfilling basic needs such as clothing, food, and shelter. The factors that play a role in determining the poverty rate in Indonesia tend to vary in each province; this is due to the diverse conditions resulting from spatial heterogeneity. However, poverty in Indonesia is not only influenced by factors from various regions but also by the conditions of the districts/cities within them. Districts/cities within a province form a spatial hierarchy structure. Therefore, in this study, the Hierarchical Linear Model (HLM), Geographically Weighted Regression (GWR), and Hierarchical and Geographically Weighted Regression (HGWR) methods were applied to determine the best model among the three methods in analyzing the factors affecting the poverty rate in Indonesia with a spatial hierarchy structure. The results of the analysis show that the HGWR method is the best model compared to HLM and GWR, as evidenced by the higher R-squared value of 0.8004 compared to HLM and GWR. Based on the HGWR model, most of the local estimators for population dependency ratio (G1), adjusted per capita expenditure (G2), and economic growth rate (G3) showed significance in provinces located in eastern Indonesia. In addition, the fixed effects and random effects estimators, namely the percentage of households without access to electricity (X1), the ratio of per capita normative consumption to net product (X2), and the percentage of households without access to clean water (X3), also have a significant influence on the poverty rate in Indonesia.