Poverty indicator until now could not be available at the village level because of limited sample coverage in the National Socio-Economic Survey that conducted by BPS. The problem could be solved by using Small Area Estimation, one of them using Spatial Empirical Best Linear Unbiased Prediction (Spatial EBLUP) that incorporate spatial aspect or location as the weights. One of goals of this research is to use a parametric bootstrap procedure to estimate the MSE and RRMSE of Spatial EBLUP method that used as a measure of its accuration compared to the direct estimation method in estimating village level poverty in Pati Regency. The response variable percentage of poverty that calculated from direct estimation, while the auxillary variable derived from the Data Collection Village Potential 2011 and Population Census 2010 i.e., characteristics of individuals, households, and areas variables that related to poverty. Spatial EBLUP for estimate village level poverty done using spatial weight matrix of customized contiguity main business field. Spatial EBLUP generates MSE (Mean Square Error) and RRMSE (Relative Root Mean Square Error) value which is smaller than the direct estimation that means the accuracy of the method is better that the direct estimation method. The visualization result from poverty mapping shows that southern part of the Pati Regency has a povert problem more serious than the northern part of the region. The estimation results with Spatial EBLUP shows that Kayen and Pucakwangi Sub-district have the most serious poverty problem in Pati Regency because most have a number of villages with the poverty classification is very high that each of three villages.