Modelling of Income Inequality in East Java Using Geographically Weighted Panel Regression

Abstract

Abstract Regression analysis is one of the statistical methods that study the relationship between response variables and predictor variables. Parameter estimates in classical linear regression produce regression coefficients that are thought to apply globally to the entire observation unit. But in fact, the existence of factors from the spatial aspect causes conditions between one location and another to be different. This spatial aspect allows the emergence of spatial heterogeneity. Geographically Weighted Regression (GWR) is a local development regression technique from ordinary regression using spatial data. In addition, in a study data is needed in a certain period of time involving cross-section data and time series or referred to as panel data. Geographically Weighted Panel Regression (GWPR) is a combination of GWR and panel data regression. The purpose of this study is to model Geographically Weighted Panel Regression using Fixed Effect Model (FEM) within estimators with adaptive bisquare kernel weight for data on income inequality (Gini ratio) in East Java Province from 2010 to 2014. In addition, to obtain factors that influence significant income inequality in each district/city of East Java Province. The results of this study indicate that the GWPR fixed effect model differs significantly in the panel data regression model, and the models produced for each location will be different from each other. Districts/cities in East Java Province have twenty-eight groups based on significant variables. The variables that significantly influences income inequality are the percentage of the poor, percentage of GDP regional in the category of fisheries forestry agriculture, percentage of GDP regional in the processing industry category, percentage of GDP regional gross fixed capital formation, per-capita GDP regional, and dependency ratio. In the GWPR model, the R2 value is 99.953%, with Root Mean Square (RMSE) is 0.0061035. While the FEM model within estimator produces an R2 value of 22.844% with RMSE is 0.1035616.