MODELING POVERTY SEVERITY INDEX IN EASTERN INDONESIA BASED ON NONPARARAMETRIC SPLINE TRUNCATED APPROACH FOR PANEL DATA

Abstrak

Poverty severity remains a critical issue in Eastern Indonesia, where rates are consistently higher than in other regions. This study examines the Poverty Severity Index (P2) using parametric panel regression and nonparametric truncated splines for panel data across 17 provinces for the period 2020 to 2024. The predictor variables include per capita expenditure, mean years of schooling, and unmet need for health services. The secondary data were obtained from the official website of Central Bureau of Statistics (BPS). The parametric FEM produces a within R² of 36.9% and an MSE of 0.00912, which provides a baseline assessment of overall trends and global relationships among variables. In parallel, the first-order truncated spline model with two knot points which produces specific-province models, achieves an R² of 99.87% and an MSE of 0.00044. This model captures detailed province-specific patterns and nonlinearities and offers additional descriptive insight into regional variations in poverty severity. Together, these complementary approaches highlight both global and local dynamics and inform policy decisions that address economic, educational, and healthcare disparities in high-poverty regions especially in Eastern Indonesia.