It is generally agreed that the primary purpose of the agricultural policy is to promote farm household income. A growing body of literature had addressed this issue, however, most of the studies were based on the aggregated data. To fill the knowledge gap, this paper utilizes Moran's I index and LISA method to empirically test the spatial dependency of the farm household income in Taiwan using the township data drawn from the 2005 Agriculture Census survey. A spatial lag and a spatial error econometric models are also estimated. Results indicate that farm household income is spatially correlated. Ignoring the spatial nature of the farm household income will lead to potential bias of the policy inference.