This article reviews the existing hierarchically spatial lag (HSLAG) model and the hierarchically spatial autoregressive error (HSEAR) model, then builds up a hierarchically spatial moving average error (HSEMA) model that incorporates the spatial moving average error and the nested random effect. In the framework of generalized moment (GMM) estimation, 18 moment conditions are derived and the corresponding estimators are proposed for the HSEMA model. Furthermore, in order to investigate the precision and finite sample properties of each estimator, Monte Carlo simulation is conducted to make comparisons among the estimation residual distribution of HSEMA, HSLAG, and HSEAR models.