Abstract: Spatial regression is one of the statistical methods that has problems of spatial dependency and heteroskedasticity. Spatial autoregressive regression (SAR) concerns only to the dependence on lag. The estimation of SAR parameters containing heteroskedasticity using the maximum likelihood estimation (MLE) method gives biased and inconsistent. The alternative method is generalized method of moments (GMM). GMM uses a combination of linear and quadratic moment functions simultaneously, so that the computation is easier than that of MLE. This study is to develop SAR model with heteroskedasticity disturbances using the GMM. The model is evaluated based on residual variance and pseudo R2. Furthermore, this method is applied to the Java’s Gross Regional Domestic Product (GRDP) data on 2017. The results showed that the district minimum wage and local revenue were significantly influence to the Java’s GRDP data in 2017. This model provides pseudo R2 value of 75.3% which means it is good enough to illustrate the diversity of Java’s GRDP in 2017.