In Japan, data on regional hourly electricity demand data by sector (residential, commercial, transport, etc.) are not publicly available; hence estimating it by bottom-up approach is important not only for electricity production and distribution companies, but also for urban planners. One of the conventional methods to estimate this is using the intensity method, in which regional electricity demand is estimated by multiplying electricity intensity (demand per floor) by floor space in each zone. Japanese Ministry of Land, Infrastructure, Transport and Tourism (MLIT) begins to publish the building stock survey from 2010, and the applicability of intensity method has dramatically raised. However, the statistics is only available at prefecture level, while more spatially finer data are required for regional electricity planning. Hence, here, we create municipality level floor space stock data by downscaling the prefecture level stock data. We conduct the downscaling by applying primal geographical methods, including the areal weighting interpolation method, the dasymetric method, the linear regression-based method. These methods have actively been discussed, and effectiveness of them, particularly the dasymetric method, has been demonstrated in many geographical literatures. On the other hand, downscaling is a recent hot topic in tics. Thus, we also apply two spatial statistical methods, including a geostatistics-based method, which captures spatially dependence spatial process in data, and our proposed geographically weighted regression-based method, which captures spatial heterogeneity by allowing parameters vary over space. The spatial statistical methods are theoretically sophisticated in that they explicitly minimize the error variances due to downscaling. However, their effectiveness in practical applications is still unclear. Thus, comparative analysis of the aforementioned methods would be important to conduct the stock data downscaling accurately. This study first applies the geographical and the spatial statistical methods for the residential floor stock data, which are available at municipality level, and accuracies of these methods are compared. The result suggests that, whereas each of the methods are fairly accurate, the spatial statistical methods possibly introduce odd results in some particular situation. Also, we discuss how the odd results can be avoided. Based on the comparative analysis, the commercial floor stock data, which is not available at municipality level, are downscaled.