关键词: 特征价格模型, 空间扩展模型, 地理加权回归, 空间异质性, 居住用地价格

Abstract: Hedonic land price models typically impose a spatial homogeneous price structure on land characteristics throughout the entire land market. However, there are increasing theoretical and empirical evidences that the marginal values of many crucial attributes of land parcels vary across space. Theoretically, localized and inelastic land supply results in spatial mismatch between demand and supply of land with certain attributes, which causes the spatial heterogeneous effects of these attributes. In this paper, we establish a series of models to evaluate the determinants of residential land price and the spatial heterogeneity of the determinants. First, we use hedonic models to diagose the determinants of residential land price. Second, we use the spatial expansion models and geographically weighted regression model (GWR) to depict the spatial instability in the impacts of land attributes. Third, we compare the prediction accuracy of the two models by predicting the land price of 10% random selected land parcels. We take Beijing as a case study and use the information of auctioned residential land parcels during 2004-2009 and GIS data of Beijing's public facilities. Based on the analysis, several conclusions are drawn as follows. 1) Spatial dependence of local residential land price and the spillover effect of local commercial land exert great effects on the residential land price, while the impact of the distance on CBD is insignificant, which indicates that the residential land market is probably local rather than global. 2) Among several public facilities, in terms of Nearest-Distance Accessibility Criteria, only the local prime elementary school, park and rail transit accessibility have a significant effect on the residential land price. 3) There is an obvious spatial pattern in the impacts of the land attributes on the land price, which is an evident signal of existence of land submarkets. 4) As the spatial expansion model imposes a fixed and definite function of spatial coordinates on the spatial heterogeneity in the marginal effect of land parcel attributes, it does not perform as well as GWR models in depicting spatial variation and prediction accuracy. GWR models perform best in explaining the land price variation, depicting spatial heterogeneity and prediction accuracy comparing to hedonic models and spatial expansion models. Also, GWR models provide a useful framework for delineating residential land submarket boundaries.

Key words: spatial expansion model, GWR, land submarkets, Beijing, residential land price, hedonic model