Spatial modeling of the open unemployment rate in West Java using eigenvector spatial filtering and spatially varying coefficients

Spatial modeling serves as a crucial approach in examining the Open Unemployment Rate (OUR) as it accommodates spatial dependence and regional heterogeneity. This study aims to model the OUR at the regency/city level in West Java Province for the 2020–2024 period using a spatial regression approach, by combining the Eigenvector Spatial Filtering (ESF) and Spatially Varying Coefficients (SVC) methods through the spmoran package in R. The ESF model is employed to reduce spatial autocorrelation in residuals by incorporating eigenvectors derived from a spatial weights matrix, while the SVC model captures local variations in the influence of explanatory variables on OUR across regions. The results reveal that the best-performing models based on Moran’s I is ESF respectively, for each year. Several variables such as Gross Regional Domestic Product (GRDP) per capita at constant 2010 prices, number of poor people, and elevation were consistently significant and influenced regional variations in OUR. Spatial visualization of model predictions indicates a concentration of high OUR values in the northern and western regions of West Java, such as Bekasi City, Karawang Regency, and Depok City, while the southern and eastern regions, such as Pangandaran and Tasikmalaya Regencies, tended to have lower OUR values. These findings underscore the importance of spatial-based policy interventions in the employment sector.