Spatial Inequality of Poverty In Lampung Province In The Perspective of Sustainable Development Using Autoregressive Spatial Approach and Spatial Error Model
Spatial regression is a development of classical regression that considers the influence of location between regions. This study aims to determine the factors that influence poverty in Lampung Province in 2024, determine the best model through a comparison of the Spatial Autoregressive (SAR) and Spatial Error Model (SEM) methods, and identify spatial patterns of poverty in Lampung Province. Model parameter estimation is carried out at each observation location using a spatial weighting matrix of Rook contiguity to capture the proximity between regions. Based on the Lagrange Multiplier test, the SAR model is identified as the best model, because it is able to capture the influence of interregional linkages through significant spatial coefficients with an AIC value of 136.3521. The SAR parameter estimation results show that the population (X_3) with a p-value of <2.2e-16 and the human development index (X_4) with a p-value of 0.006990, with a significance level or α <0.10, have a significant effect on the number of poor people in Lampung Province, while gross regional domestic product (X_1) and TPAK (X_2) have no significant effect. The Local Indicators of Spatial Association (LISA) analysis also revealed the existence of High-High and Low-Low poverty clusters indicating spatial grouping in several regencies/cities in Lampung Province. However, after SAR modeling, the residual map of the SAR model shows a scattered spatial pattern and does not form a geographic cluster, indicating that the influence between regions has been successfully captured by the model. This finding confirms that poverty in Lampung Province is not only influenced by the internal characteristics of the region, but is also influenced by the conditions of the surrounding areas. Therefore, policy actions need to consider the spatial effects between regions so that poverty alleviation programs are more targeted and effective.