SAR probit regression modeling using the fisher scoring approach: A case study of poverty levels on the island of Sumatra

The probit model is commonly used to study categorical response data. However, failing to account for spatial autocorrelation factors between regions can lead to inconsistent and biased parameter estimation results. This study focuses on examining the parameter estimation of the Spatial Autoregressive (SAR) Probit model through the Maximum Likelihood Estimation (MLE) method using the Fisher Scoring numerical iteration scheme. The model was implemented on poverty data across 131 regencies/cities in the Sumatra region for 2022. Empirical findings indicate that the GRDP growth rate, open unemployment rate, per capita expenditure, and expected years of schooling significantly affect the poverty rate. The results of this model development provide a prediction accuracy of 83.97%. This achievement is superior to that of the RIS Simulator technique, which only yields an accuracy of 74.05%. These results emphasize the advantages of the efficiency and accuracy of the Fisher Scoring approach in representing spatial dependencies in the poverty phenomenon in Sumatra.