Non-Stationary Semivariogram Analysis Using Real Estate Transaction Data

The development of information and communication technologies (Internet, databases ...) reduces the barriers of investment. Consequently, investors now have a wealth of opportunities available for diversifying their real estate portfolio geographically. However, neighborhoods residential properties tend to have similar price evolution because they have the same structural characteristics and share location amenities. The previous studies often confirm a degree of spatial autocorrelation (positive or very positive autocorrelation) between neighboring properties. Hence, the real estate diversification between predefined regions, based on administrative boundary (arrondissement) is rarely optimal. Differences from the administrative segmentation, this study analyzes the relevance of a new segmentation of Paris housing markets that could improve the geographical diversification performance. By applying the spatial econometrics techniques based on the notaryís data of 35206 apartmentsí transaction in Paris in 2007, we attempt to use residual spatial autocorrelation information to redefine new market segmentation. This geographical boundary allows the properties to determine their submarket structure and to eliminate the spatial autocorrelation problem between submarket. We find a low spatial autocorrelation of properties belonging to different submarkets. According to this study, the diversified portfolios based on this structure show probably more efficient than the previous literatures established on the traditional administrative segmentation.