In regression analysis, the independence assumption of the error terms is often violated when working with spatial data. The 2023 poverty incidence data across regencies/municipalities on Sulawesi Island indicate the presence of spatial autocorrelation. This study aims to compare the performance of classical regression, spatial autoregressive model (SAR), and spatial error model (SEM) in modeling poverty incidence on the island. The regency/municipality-level data used in the study is secondary data published by BPS-Statistics Indonesia. The findings reveal that the SEM model provides more accurate parameter estimates compared to classical regression and SAR model. Factors that have a significant influence on the poverty incidence (Y) in a regency/municipality are life expectancy (X1), expenditure per capita (X2), and the error terms for the nearest neighboring regions (λ).